Parte I
Conhecer
Capítulo 1
O SIMUT em uma página
Este capítulo diz o que o SIMUT faz, para quem ele serve e o que ele não é, mostra os três aparelhos lado a lado e ensina a ler o resto do manual. É para todo leitor, antes de qualquer outro capítulo.
O que o SIMUT faz
-O SIMUT, Sistema Integrado de Monitoramento Universal e Telemetria, é um firmware para o Raspberry Pi Pico W e o Pico 2 W que transforma a placa num aparelho de monitoramento e telemetria: hoje de temperatura, umidade e pressão, e de qualquer grandeza para a qual se escreva um driver. Um aparelho SIMUT:
+O SIMUT, Sistema Integrado de Monitoramento Universal e Telemetria, é um firmware para o Raspberry Pi Pico W e o Pico 2 W que transforma a placa num aparelho de monitoramento e telemetria: hoje de temperatura, umidade e pressão, e de outras grandezas até o limite da tabela de canais, que tem quatro linhas das oito que comporta, cada grandeza com o seu driver. Um aparelho SIMUT:
- lê até 16 sensores dos tipos DS18B20, DHT22, BME280 e BMP280, ligados aos pinos GP0 a GP15 (capítulo 2, capítulo 6);
- grava o histórico no próprio aparelho, um registro por intervalo, a partir de 1 min. Com 11 grandezas e um registro por minuto, cabem cerca de 116 dias (medido em 31/07/2026; capítulo 15); @@ -9273,7 +9273,7 @@
A tela de atualização
As telas de boot
-O Ângulo desenha o boot numa tela só: a marca SIMUT, Monitor de temperatura e umidade (Temperature and humidity monitor), as últimas linhas do registro de boot, com a etapa atual em letra maior, e, embaixo, a versão e o endereço MAC. As linhas, a ordem, o tempo de cada etapa e o gesto do ponto de acesso são do núcleo 0 e iguais nos dois painéis (As telas de boot); o Ângulo só os desenha.
O Ângulo desenha o boot numa tela só: a marca SIMUT, Monitoramento universal e telemetria (Universal monitoring and telemetry; até 07/11/2026 dizia “Monitor de temperatura e umidade”), as últimas linhas do registro de boot, com a etapa atual em letra maior, e, embaixo, a versão e o endereço MAC. As linhas, a ordem, o tempo de cada etapa e o gesto do ponto de acesso são do núcleo 0 e iguais nos dois painéis (As telas de boot); o Ângulo só os desenha.
Enquanto o aparelho sobe, só um botão responde ao toque: Pular a espera (Skip the wait), que aparece quando o núcleo 0 oferece e faz o boot seguir. A volta por inatividade não vale no boot.
O gesto do ponto de acesso é do núcleo 0: ele abre uma janela de 3,5 s e pede que você segure a tela por 3 s. Durante a janela, o Ângulo troca as linhas do boot por Segure para o modo de configuração (Hold to enter configuration mode) e uma barra que enche com o tempo segurado.
O que a captura e o espelho mostram
diff --git a/docs/_manual/cap01-o-simut.md b/docs/_manual/cap01-o-simut.md index a0071e52..d4a97fed 100644 --- a/docs/_manual/cap01-o-simut.md +++ b/docs/_manual/cap01-o-simut.md @@ -4,7 +4,7 @@ Este capítulo diz o que o SIMUT faz, para quem ele serve e o que ele não é, m ## O que o SIMUT faz {#cap-01-o-que} -O SIMUT, Sistema Integrado de Monitoramento Universal e Telemetria, é um firmware para o Raspberry Pi Pico W e o Pico 2 W que transforma a placa num aparelho de monitoramento e telemetria: hoje de temperatura, umidade e pressão, e de qualquer grandeza para a qual se escreva um driver. Um aparelho SIMUT: +O SIMUT, Sistema Integrado de Monitoramento Universal e Telemetria, é um firmware para o Raspberry Pi Pico W e o Pico 2 W que transforma a placa num aparelho de monitoramento e telemetria: hoje de temperatura, umidade e pressão, e de outras grandezas até o limite da tabela de canais, que tem quatro linhas das oito que comporta, cada grandeza com o seu driver. Um aparelho SIMUT: - **lê até 16 sensores** dos tipos DS18B20, DHT22, BME280 e BMP280, ligados aos pinos GP0 a GP15 ([capítulo 2](#cap-02), [capítulo 6](#cap-06)); - **grava o histórico no próprio aparelho**, um registro por intervalo, a partir de 1 min. Com 11 grandezas e um registro por minuto, cabem cerca de 116 dias (medido em 31/07/2026; [capítulo 15](#cap-15)); diff --git a/docs/_manual/cap11a-angulo.md b/docs/_manual/cap11a-angulo.md index 02109a5c..80a03c24 100644 --- a/docs/_manual/cap11a-angulo.md +++ b/docs/_manual/cap11a-angulo.md @@ -657,7 +657,7 @@ Uma atualização interrompida: o envio parou antes do fim. ### As telas de boot {#cap-11a-boot} -O Ângulo desenha o boot numa tela só: a marca `SIMUT`, **Monitor de temperatura e umidade** (*Temperature and humidity monitor*), as últimas linhas do registro de boot, com a etapa atual em letra maior, e, embaixo, a versão e o endereço MAC. As linhas, a ordem, o tempo de cada etapa e o gesto do ponto de acesso são do núcleo 0 e iguais nos dois painéis ([As telas de boot](#cap-11-boot)); o Ângulo só os desenha. +O Ângulo desenha o boot numa tela só: a marca `SIMUT`, **Monitoramento universal e telemetria** (*Universal monitoring and telemetry*; até 07/11/2026 dizia "Monitor de temperatura e umidade"), as últimas linhas do registro de boot, com a etapa atual em letra maior, e, embaixo, a versão e o endereço MAC. As linhas, a ordem, o tempo de cada etapa e o gesto do ponto de acesso são do núcleo 0 e iguais nos dois painéis ([As telas de boot](#cap-11-boot)); o Ângulo só os desenha. Enquanto o aparelho sobe, só um botão responde ao toque: **Pular a espera** (*Skip the wait*), que aparece quando o núcleo 0 oferece e faz o boot seguir. A volta por inatividade não vale no boot. diff --git a/platformio.ini b/platformio.ini index dbc6b3a5..d5dc3b98 100644 --- a/platformio.ini +++ b/platformio.ini @@ -1,6 +1,5 @@ ; ============================================================================ ; SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria -; Integrated Universal Monitoring and Telemetry System ; Target: Raspberry Pi Pico W (RP2040), framework Arduino-Pico (earlephilhower) ; ============================================================================ diff --git a/src/AppManager.h b/src/AppManager.h index 2191640f..59c79da7 100644 --- a/src/AppManager.h +++ b/src/AppManager.h @@ -7,7 +7,6 @@ * alarm monitoring, sensor calibration, and NTP time correction. * * @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria - * SIMUT — Integrated Universal Monitoring and Telemetry System * @target Raspberry Pi Pico W (RP2040) — Arduino Framework * @author Ângelo Moisés Alves * @license MIT License diff --git a/src/AppManager_Boot.cpp b/src/AppManager_Boot.cpp index ad2e7942..34e3d98e 100644 --- a/src/AppManager_Boot.cpp +++ b/src/AppManager_Boot.cpp @@ -2,7 +2,6 @@ * @file AppManager_Boot.cpp * @brief Boot sequence: filesystem, sensors, network, web server initialization. * @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria - * SIMUT — Integrated Universal Monitoring and Telemetry System * @target Raspberry Pi Pico W (RP2040) — Arduino Framework * @author Ângelo Moisés Alves * @license MIT License diff --git a/src/AppManager_CmdHandlers.cpp b/src/AppManager_CmdHandlers.cpp index 1195ea99..905aa424 100644 --- a/src/AppManager_CmdHandlers.cpp +++ b/src/AppManager_CmdHandlers.cpp @@ -6,7 +6,6 @@ * readability (executeCommand reduced from 952 to ~600 lines). * * @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria - * SIMUT — Integrated Universal Monitoring and Telemetry System * @target Raspberry Pi Pico W (RP2040) — Arduino Framework * @author Ângelo Moisés Alves * @license MIT License diff --git a/src/AppManager_Commands.cpp b/src/AppManager_Commands.cpp index 805f290c..075680ed 100644 --- a/src/AppManager_Commands.cpp +++ b/src/AppManager_Commands.cpp @@ -2,7 +2,6 @@ * @file AppManager_Commands.cpp * @brief CLI command execution (40+ cases across config, sensors, network, users). * @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria - * SIMUT — Integrated Universal Monitoring and Telemetry System * @target Raspberry Pi Pico W (RP2040) — Arduino Framework * @author Ângelo Moisés Alves * @license MIT License diff --git a/src/AppManager_Core.cpp b/src/AppManager_Core.cpp index 62f10280..7c10b5f2 100644 --- a/src/AppManager_Core.cpp +++ b/src/AppManager_Core.cpp @@ -2,7 +2,6 @@ * @file AppManager_Core.cpp * @brief Core infrastructure: constructor, display health, time sync, user interaction. * @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria - * SIMUT — Integrated Universal Monitoring and Telemetry System * @target Raspberry Pi Pico W (RP2040) — Arduino Framework * @author Ângelo Moisés Alves * @license MIT License diff --git a/src/AppManager_Events.cpp b/src/AppManager_Events.cpp index d70889a1..f8f87cf7 100644 --- a/src/AppManager_Events.cpp +++ b/src/AppManager_Events.cpp @@ -2,7 +2,6 @@ * @file AppManager_Events.cpp * @brief UI event dispatch: touch, graph, calendar, settings, alarms. * @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria - * SIMUT — Integrated Universal Monitoring and Telemetry System * @target Raspberry Pi Pico W (RP2040) — Arduino Framework * @author Ângelo Moisés Alves * @license MIT License diff --git a/src/AppManager_Graph.cpp b/src/AppManager_Graph.cpp index 40740ef3..7160a975 100644 --- a/src/AppManager_Graph.cpp +++ b/src/AppManager_Graph.cpp @@ -2,7 +2,6 @@ * @file AppManager_Graph.cpp * @brief Graph rendering from binary history: cache, render, preload, append. * @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria - * SIMUT — Integrated Universal Monitoring and Telemetry System * @target Raspberry Pi Pico W (RP2040) — Arduino Framework * @author Ângelo Moisés Alves * @license MIT License diff --git a/src/AppManager_HistoryAlarm.cpp b/src/AppManager_HistoryAlarm.cpp index ac686c91..bf1d5088 100644 --- a/src/AppManager_HistoryAlarm.cpp +++ b/src/AppManager_HistoryAlarm.cpp @@ -2,7 +2,6 @@ * @file AppManager_HistoryAlarm.cpp * @brief History logging, min/max computation, alarm monitoring, live display updates. * @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria - * SIMUT — Integrated Universal Monitoring and Telemetry System * @target Raspberry Pi Pico W (RP2040) — Arduino Framework * @author Ângelo Moisés Alves * @license MIT License diff --git a/src/AppManager_Loop.cpp b/src/AppManager_Loop.cpp index 512f900b..0d2a181c 100644 --- a/src/AppManager_Loop.cpp +++ b/src/AppManager_Loop.cpp @@ -2,7 +2,6 @@ * @file AppManager_Loop.cpp * @brief Main loop with priority-based task scheduling. * @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria - * SIMUT — Integrated Universal Monitoring and Telemetry System * @target Raspberry Pi Pico W (RP2040) — Arduino Framework * @author Ângelo Moisés Alves * @license MIT License diff --git a/src/AppManager_Sensors.cpp b/src/AppManager_Sensors.cpp index 6555396b..5e05ba07 100644 --- a/src/AppManager_Sensors.cpp +++ b/src/AppManager_Sensors.cpp @@ -2,7 +2,6 @@ * @file AppManager_Sensors.cpp * @brief Sensor health: auto-heal, background scan, calibration. * @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria - * SIMUT — Integrated Universal Monitoring and Telemetry System * @target Raspberry Pi Pico W (RP2040) — Arduino Framework * @author Ângelo Moisés Alves * @license MIT License diff --git a/src/B64Decode.h b/src/B64Decode.h index 11044066..d84af1c7 100644 --- a/src/B64Decode.h +++ b/src/B64Decode.h @@ -8,7 +8,6 @@ * credential. * * @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria - * SIMUT — Integrated Universal Monitoring and Telemetry System * @target Raspberry Pi Pico W (RP2040) — Arduino Framework * @author Ângelo Moisés Alves * @license MIT License diff --git a/src/BluetoothManager.cpp b/src/BluetoothManager.cpp index 02b53a2e..e185ee8d 100644 --- a/src/BluetoothManager.cpp +++ b/src/BluetoothManager.cpp @@ -6,7 +6,6 @@ * All output methods are gated by authentication status. * * @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria - * SIMUT — Integrated Universal Monitoring and Telemetry System * @target Raspberry Pi Pico W (RP2040) — Arduino Framework * @author Ângelo Moisés Alves * @license MIT License diff --git a/src/CommandManager.cpp b/src/CommandManager.cpp index 267bdbbf..796507d0 100644 --- a/src/CommandManager.cpp +++ b/src/CommandManager.cpp @@ -6,7 +6,6 @@ * system info, and log entries. Supports dual-buffer USB/BT processing. * * @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria - * SIMUT — Integrated Universal Monitoring and Telemetry System * @target Raspberry Pi Pico W (RP2040) — Arduino Framework * @author Ângelo Moisés Alves * @license MIT License diff --git a/src/CommandManager.h b/src/CommandManager.h index e0621c19..c75d04c9 100644 --- a/src/CommandManager.h +++ b/src/CommandManager.h @@ -6,7 +6,6 @@ * both USB Serial and authenticated Bluetooth sessions. * * @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria - * SIMUT — Integrated Universal Monitoring and Telemetry System * @target Raspberry Pi Pico W (RP2040) — Arduino Framework * @author Ângelo Moisés Alves * @license MIT License diff --git a/src/HaDiscovery.h b/src/HaDiscovery.h index 5f0ceb29..6b2d3325 100644 --- a/src/HaDiscovery.h +++ b/src/HaDiscovery.h @@ -13,7 +13,6 @@ * the discovery message is a promise about keys that formatter emits. * * @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria - * SIMUT — Integrated Universal Monitoring and Telemetry System * @target Raspberry Pi Pico W (RP2040) — Arduino Framework * @author Ângelo Moisés Alves * @license MIT License diff --git a/src/HelpLicenseEN.h b/src/HelpLicenseEN.h index 105b7034..6fd93175 100644 --- a/src/HelpLicenseEN.h +++ b/src/HelpLicenseEN.h @@ -5,7 +5,6 @@ * now embedded in firmware. Not accessible to user via /files. * PT versions come from /lang/language_.lng (.lng parser).
* @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria
- * SIMUT — Integrated Universal Monitoring and Telemetry System
* @author Ângelo Moisés Alves
* @license MIT
*/
diff --git a/src/LogPolicy.cpp b/src/LogPolicy.cpp
index b654d739..a57173f7 100644
--- a/src/LogPolicy.cpp
+++ b/src/LogPolicy.cpp
@@ -3,7 +3,6 @@
* @brief Routing table and state machine for the edge-triggered log filter.
*
* @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria
- * SIMUT — Integrated Universal Monitoring and Telemetry System
* @author Ângelo Moisés Alves
* @license MIT License
*/
diff --git a/src/LogPolicy.h b/src/LogPolicy.h
index 943959e8..0e392e61 100644
--- a/src/LogPolicy.h
+++ b/src/LogPolicy.h
@@ -35,7 +35,6 @@
* parameter for the same reason.
*
* @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria
- * SIMUT — Integrated Universal Monitoring and Telemetry System
* @author Ângelo Moisés Alves
* @license MIT License
*/
diff --git a/src/PromMetrics.h b/src/PromMetrics.h
index 6696db4d..33f65724 100644
--- a/src/PromMetrics.h
+++ b/src/PromMetrics.h
@@ -12,7 +12,6 @@
* knows how a line is spelled.
*
* @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria
- * SIMUT — Integrated Universal Monitoring and Telemetry System
* @target Raspberry Pi Pico W (RP2040) — Arduino Framework
* @author Ângelo Moisés Alves
* @license MIT License
diff --git a/src/Syslog5424.h b/src/Syslog5424.h
index 618446a7..1ca2f1af 100644
--- a/src/Syslog5424.h
+++ b/src/Syslog5424.h
@@ -28,7 +28,6 @@
* Every structured field is sanitized to 33..126, empty → NILVALUE.
*
* @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria
- * SIMUT — Integrated Universal Monitoring and Telemetry System
* @target Raspberry Pi Pico W (RP2040) — Arduino Framework
* @author Ângelo Moisés Alves
* @license MIT License
diff --git a/src/SyslogManager.h b/src/SyslogManager.h
index c171b1db..dad650c9 100644
--- a/src/SyslogManager.h
+++ b/src/SyslogManager.h
@@ -24,7 +24,6 @@
* must not log from inside enqueue (it holds _logMutex) — it does not.
*
* @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria
- * SIMUT — Integrated Universal Monitoring and Telemetry System
* @target Raspberry Pi Pico W (RP2040) — Arduino Framework
* @author Ângelo Moisés Alves
* @license MIT License
diff --git a/src/SystemDefs.h b/src/SystemDefs.h
index 4d898863..a97d2915 100644
--- a/src/SystemDefs.h
+++ b/src/SystemDefs.h
@@ -17,7 +17,6 @@
* · SystemDefs_Validate.h — input validators
*
* @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria
- * SIMUT — Integrated Universal Monitoring and Telemetry System
* @target Raspberry Pi Pico W (RP2040) — Arduino Framework
* @author Ângelo Moisés Alves
* @license MIT License
diff --git a/src/SystemDefs_Cli.h b/src/SystemDefs_Cli.h
index 1f08aa5c..e15436bc 100644
--- a/src/SystemDefs_Cli.h
+++ b/src/SystemDefs_Cli.h
@@ -7,7 +7,6 @@
* SystemDefs.h (facade).
*
* @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria
- * SIMUT — Integrated Universal Monitoring and Telemetry System
* @author Ângelo Moisés Alves
* @license MIT License
*/
diff --git a/src/SystemDefs_Limits.h b/src/SystemDefs_Limits.h
index 467d1e19..924eac41 100644
--- a/src/SystemDefs_Limits.h
+++ b/src/SystemDefs_Limits.h
@@ -6,7 +6,6 @@
* Sub-header of SystemDefs.h (facade).
*
* @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria
- * SIMUT — Integrated Universal Monitoring and Telemetry System
* @author Ângelo Moisés Alves
* @license MIT License
*/
diff --git a/src/SystemDefs_Logging.h b/src/SystemDefs_Logging.h
index db70f94c..a3976c69 100644
--- a/src/SystemDefs_Logging.h
+++ b/src/SystemDefs_Logging.h
@@ -6,7 +6,6 @@
* binary record. Sub-header of SystemDefs.h (facade).
*
* @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria
- * SIMUT — Integrated Universal Monitoring and Telemetry System
* @author Ângelo Moisés Alves
* @license MIT License
*/
diff --git a/src/SystemDefs_Network.h b/src/SystemDefs_Network.h
index 2a22d66e..f66f8eda 100644
--- a/src/SystemDefs_Network.h
+++ b/src/SystemDefs_Network.h
@@ -6,7 +6,6 @@
* telemetry cursor coalesce. Sub-header of SystemDefs.h (facade).
*
* @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria
- * SIMUT — Integrated Universal Monitoring and Telemetry System
* @author Ângelo Moisés Alves
* @license MIT License
*/
diff --git a/src/SystemDefs_Records.h b/src/SystemDefs_Records.h
index 2a7dfc84..54d4f95c 100644
--- a/src/SystemDefs_Records.h
+++ b/src/SystemDefs_Records.h
@@ -9,7 +9,6 @@
* Sub-header of SystemDefs.h (facade).
*
* @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria
- * SIMUT — Integrated Universal Monitoring and Telemetry System
* @author Ângelo Moisés Alves
* @license MIT License
*/
diff --git a/src/SystemDefs_Reserved.h b/src/SystemDefs_Reserved.h
index 0c1831b0..2bbb245e 100644
--- a/src/SystemDefs_Reserved.h
+++ b/src/SystemDefs_Reserved.h
@@ -27,7 +27,6 @@
* for `reserved[` before trusting this comment.
*
* @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria
- * SIMUT — Integrated Universal Monitoring and Telemetry System
* @author Ângelo Moisés Alves
* @license MIT License
*/
diff --git a/src/SystemDefs_Time.h b/src/SystemDefs_Time.h
index 7e8d9982..4816f270 100644
--- a/src/SystemDefs_Time.h
+++ b/src/SystemDefs_Time.h
@@ -5,7 +5,6 @@
* timeSince/timeRemaining. Sub-header of SystemDefs.h (facade).
*
* @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria
- * SIMUT — Integrated Universal Monitoring and Telemetry System
* @author Ângelo Moisés Alves
* @license MIT License
*/
diff --git a/src/SystemDefs_Validate.h b/src/SystemDefs_Validate.h
index d1cf7a3d..400abfe4 100644
--- a/src/SystemDefs_Validate.h
+++ b/src/SystemDefs_Validate.h
@@ -7,7 +7,6 @@
* Sub-header of SystemDefs.h (facade).
*
* @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria
- * SIMUT — Integrated Universal Monitoring and Telemetry System
* @author Ângelo Moisés Alves
* @license MIT License
*/
diff --git a/src/SystemUtils.cpp b/src/SystemUtils.cpp
index 6fc8e53d..8fa2b925 100644
--- a/src/SystemUtils.cpp
+++ b/src/SystemUtils.cpp
@@ -6,7 +6,6 @@
* history filename format validation (YYYYMMDD.h5).
*
* @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria
- * SIMUT — Integrated Universal Monitoring and Telemetry System
* @target Raspberry Pi Pico W (RP2040) — Arduino Framework
* @author Ângelo Moisés Alves
* @license MIT License
diff --git a/src/TelemetryManager.cpp b/src/TelemetryManager.cpp
index d8244f55..3f177291 100644
--- a/src/TelemetryManager.cpp
+++ b/src/TelemetryManager.cpp
@@ -8,7 +8,6 @@
* format builders (JSON, CSV, custom template).
*
* @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria
- * SIMUT — Integrated Universal Monitoring and Telemetry System
* @target Raspberry Pi Pico W (RP2040) — Arduino Framework
* @author Ângelo Moisés Alves
* @license MIT License
diff --git a/src/TelemetryManager.h b/src/TelemetryManager.h
index 4972fd5a..1e966e1f 100644
--- a/src/TelemetryManager.h
+++ b/src/TelemetryManager.h
@@ -7,7 +7,6 @@
* count estimator for dashboard display.
*
* @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria
- * SIMUT — Integrated Universal Monitoring and Telemetry System
* @target Raspberry Pi Pico W (RP2040) — Arduino Framework
* @author Ângelo Moisés Alves
* @license MIT License
diff --git a/src/TftWithOffset.h b/src/TftWithOffset.h
index c6055247..b8d097e7 100644
--- a/src/TftWithOffset.h
+++ b/src/TftWithOffset.h
@@ -12,7 +12,6 @@
* which has been the case in the official library since 2019.
*
* @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria
- * SIMUT — Integrated Universal Monitoring and Telemetry System
* @target Raspberry Pi Pico W (RP2040) — Arduino Framework
* @author Ângelo Moisés Alves
* @license MIT License
diff --git a/src/Themes.h b/src/Themes.h
index de6cc23f..bd11c391 100644
--- a/src/Themes.h
+++ b/src/Themes.h
@@ -6,7 +6,6 @@
* persistent selection stored in SystemConfig.
*
* @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria
- * SIMUT — Integrated Universal Monitoring and Telemetry System
* @target Raspberry Pi Pico W (RP2040) — Arduino Framework
* @author Ângelo Moisés Alves
* @license MIT License
diff --git a/src/TouchPriority.h b/src/TouchPriority.h
index 1a3227c4..2961f079 100644
--- a/src/TouchPriority.h
+++ b/src/TouchPriority.h
@@ -19,7 +19,6 @@
* that it is pointer-only reads — naturally atomic on ARM 32-bit.
*
* @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria
- * SIMUT — Integrated Universal Monitoring and Telemetry System
* @target Raspberry Pi Pico W (RP2040) — Arduino Framework
* @author Ângelo Moisés Alves
* @license MIT License
diff --git a/src/WebManager_Metrics.cpp b/src/WebManager_Metrics.cpp
index 25a65c6d..6fd96c19 100644
--- a/src/WebManager_Metrics.cpp
+++ b/src/WebManager_Metrics.cpp
@@ -15,7 +15,6 @@
* audit-logged: at one line per scrape interval they would bury the log.
*
* @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria
- * SIMUT — Integrated Universal Monitoring and Telemetry System
* @target Raspberry Pi Pico W (RP2040) — Arduino Framework
* @author Ângelo Moisés Alves
* @license MIT License
diff --git a/src/main.cpp b/src/main.cpp
index f0326973..cd6cb730 100644
--- a/src/main.cpp
+++ b/src/main.cpp
@@ -7,7 +7,6 @@
* Cross-core health is monitored separately by LogManager.
*
* @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria
- * SIMUT — Integrated Universal Monitoring and Telemetry System
* @target Raspberry Pi Pico W (RP2040) — Arduino Framework
* @author Ângelo Moisés Alves
* @license MIT License
diff --git a/src/ui/original/Themes.cpp b/src/ui/original/Themes.cpp
index 5a7cfb7d..c61edeeb 100644
--- a/src/ui/original/Themes.cpp
+++ b/src/ui/original/Themes.cpp
@@ -8,7 +8,6 @@
* for maintenance (new colors/adjustments should be replicated in all).
*
* @project SIMUT — Sistema Integrado de Monitoramento Universal e Telemetria
- * SIMUT — Integrated Universal Monitoring and Telemetry System
* @target Raspberry Pi Pico W (RP2040) — Arduino Framework
* @author Ângelo Moisés Alves
* @license MIT License
diff --git a/tools/flash_budget.json b/tools/flash_budget.json
index f214c647..cd2e5fa8 100644
--- a/tools/flash_budget.json
+++ b/tools/flash_budget.json
@@ -551,7 +551,7 @@
"budget still holds, and goes to measured: 1,028,628 B."
],
"ceiling": 1044480,
- "measured_on": "2026-10-05 chore/v2.11.1: SIMUT_VERSION 2.11.1, the only source change. All seven rebuilt here. The version has as many characters as 2.11.0, but the login page carries it, and its gzip came out 5 B smaller in every image: measured and .bin 8 B under the record on release, test, asserts and pico2_w_release; alpha, air and test_https equal, their alignment absorbing it. Budgets unchanged. 2026-10-12 feat/lan-switch: pico_wire joins (the SIMUT Wire, a plain Pico with the network on an ENC28J60, no driver yet), measured 587,980 B and a .bin of 600,016 B; the seven others are byte for byte what main 2a391da builds. Then twelve sensor slots (-DMAX_SENSORS=12, GP12-GP15 are the module's): 588,020 B and a .bin of 600,056 B, +40 B for clearing the slots past the twelfth on every config load; RAM 102,416 -> 102,064 B. Then eleven (the module's INT moved to GP11, 2026-10-12): 587,972 B and a .bin of 600,008 B, RAM 101,980 B, measured on 28a54c8. 2026-10-13 feat/wire-driver: the wired driver takes pico_wire to 597,448 B and a .bin of 609,480 B (RAM 104,680 B); the seven Wi-Fi images stay byte for byte what dd7efdc builds until the top-panel fix, which grows them by 32 to 64 B (release, test, test_https and asserts +48, alpha +64, air +32, pico2_w_release +56; .bin +48 on release, test and asserts, +56 on pico2_w_release, unchanged on alpha and air). All eight measured at 4912762's tree plus this change. The review's fixes (module before the sensors, the static address checks, the MAC before the start, syslog and the OTA teardown on the cable, the log latches): pico_wire 598,560 B (+1,112 B), .bin 610,592 B, RAM 104,688 B; the seven Wi-Fi images compile none of it. 2026-10-13 feat/wire-surface: pico_wiretft joins (the SIMUT Wire with the TFT panel and touch): 780,628 B and a .bin of 792,664 B (RAM 109,784 B), built on the driver's tree plus the manifest entry and no change to src/. The eight others keep their records: no flag and no source of theirs moved. Then the surface a wired build loses with the radio (the web pages, the console, the panels and the third-party list): pico_wire 595,632 B and a .bin of 607,664 B (RAM 104,684 B), pico_wiretft 777,428 B and a .bin of 789,464 B (RAM 109,788 B), both budgets brought down to them; the seven radio images byte for byte the same at each of the branch's commits, compared by sha256 (pico_w_release 8b565f66 throughout). Then the captive-portal redirect and probes, the setup AP's, compiled out of the wired builds: pico_wire 595,016 B (.bin 607,048), pico_wiretft 776,820 B (.bin 788,856), RAM unchanged; budgets brought down again. Then four log codes of the cable's own (529-532) with their three rows in the edge filter, compiled into the wired builds only: pico_wire 595,152 B (+136, .bin 607,184), pico_wiretft 776,956 B (+136, .bin 788,992); the seven radio images unchanged. 2026-10-13 feat/net-boundary: the wired link generic over LwipIntfDev (the chip in a device header, src/net/Enc28j60Device.h) and a module stopped by an update that did not happen started again by the retry: pico_wire 595,208 B (+56, .bin 607,240), pico_wiretft 777,004 B (+48, .bin 789,040), RAM unchanged. The move alone is byte for byte neutral (pico_wire sha256 3edd1440, pico_wiretft e58d65e9, before WiredLink::chipName( ) existed); the growth is the module-fail message built from the driver's name (a second concatenation in startWired( ), plus chipName( ) and started( ) now reached) and the restart path in wiredLost( ), the only two function bodies that differ in the normalized disassembly of both images; the seven radio images are byte for byte the same. Then the log line of that path names what happened (\"module stopped\", where it said \"link down\"): +48 B on both, pico_wire 595,256 B (.bin 607,288), pico_wiretft 777,052 B (.bin 789,088); the seven radio images unchanged. Then the review of that change: the stopped module is looked for first in updateWired( ), in every state, because a unit still on its way up (NET_CONNECTING: booted with the cable out, or a lease pending) never reached wiredLost( ) and was never started again: +40 B on both, pico_wire 595,296 B (.bin 607,328), pico_wiretft 777,092 B (.bin 789,128), RAM unchanged; updateWired( ) and wiredLost( ) the only two function bodies that differ in the normalized disassembly of both images. The seven radio images are byte for byte the same, pico_w_air included: its radio power-down moved into net::powerDownForSleep( ) (src/net/NetLink.h) without moving a byte. 2026-10-13 feat/net-radio-unit: the radio's state machine moved from NetworkManager.cpp into its own unit, src/net/NetworkManager_Cyw43.cpp, which the wired images do not build, and NetworkManager.cpp became the host of what every network needs, reaching the link through three methods each unit defines. The wired images lose the radio code they compiled dead: pico_wire 590,740 B (-4,556, .bin 602,776), pico_wiretft 772,852 B (-4,240, .bin 784,888), both budgets brought down. The radio images pay for the split: release +48 B (1,021,180, .bin 1,033,212), asserts +40 (1,023,468, .bin 1,035,500), pico2_w_release +120 (997,328, .bin 997,352), test_https -8 (1,030,692), test, alpha and air the same; the normalized disassembly of all seven differs in four bodies only: begin( ) and update( ) (now the host's), handleConnecting( ) and getRssi( ), which call resetReconnectLadder( ) and isConnected( ) across the unit boundary where the compiler inlined them before. Inlining the two helpers in the header instead was measured and grew the radio images more (release +88, test +56, asserts +80). 2026-10-13 fix/radio-link-in-ntp-wait: the radio watches its link inside the NTP wait, one check in its updateLinkFirst( ) and NET_READY's loss in a helper that logs it, wifiLost( ). The seven radio images grow: release +72 B (1,021,252, .bin 1,033,284), test +80 (1,025,788, .bin 1,037,820, under its budget), test_https +80 (1,030,772), asserts +80 (1,023,548, .bin 1,035,580), alpha +88 (975,060, under its budget, the .bin the same), air +80 (1,018,336, under its budget, the .bin the same), pico2_w_release +72 (997,400, .bin 997,424); the two wired images are byte for byte the same. With the log at each of the two callers instead of in the helper, release grew +96 B. 2026-10-13 fix/silent-radio-no-blind-join (#266): a reconnect scan whose start the radio did not answer, and that never finished, is not escaped by a blind join; two in a row past the first half hour restart the radio by safeReboot( ). The radio images grow: release +376 B (1,021,628, .bin 1,033,660), test +352 (1,026,140, .bin 1,038,172), test_https +352 (1,031,124), asserts +384 (1,023,932, .bin 1,035,964), alpha +280 (975,340) and air +280 (1,018,616), both .bin the same, pico2_w_release +376 (997,776, .bin 997,800). The wired images carry the two new members of the shared header and the host's reset of one: pico_wire the same size (590,740), pico_wiretft +16 (772,868, .bin 784,904). Writing the two log lines with one snprintf each instead of the String concatenations was measured: 16 B less on release, not worth the buffer. 2026-10-14 fix/link-health-in-ntp-wait: getRssi( ) answers the level of any link that is up, the NTP wait included, on the radio and on the cable (isLinkUp( ) where it read isConnected( )). The RP2040 images are the same size; pico2_w_release +8 B (997,784, .bin 997,808). With it, the Air's FLUSH waits for NTP 10 s at most (AIR_FLUSH_NTP_WAIT_MS, airFlushNetLost( )): pico_w_air +48 B (1,018,664, the .bin the same), the other eight byte for byte the same. 2026-10-14 fix/web-short-write: override 2m (webserver_short_write.patch), a write cut short ends the response and closes the client as its real type (stopClient( ), so over TLS too), and the count read out (metr.cgw, /metrics, and `show net status` with the TCP segment pool and tcp_write's memory failures). Every image grows: release +512 B (1,022,140, .bin 1,034,172), test +480 (1,026,620, .bin 1,038,652), test_https +504 (1,031,628), asserts +504 (1,024,436, .bin 1,036,468), alpha +344 (975,684) and air +352 (1,019,016), both .bin the same, pico2_w_release +520 (998,304, .bin 998,328), pico_wire +480 (591,220, .bin 603,256), pico_wiretft +504 (773,372, .bin 785,408). Of the release's 512 B, 394 are sized symbols (the status handler 72, _writeAll( ) and _shortWrite( ) 78, the console 56, the metrics 32, the three _currentClientStop( ) 32, the framework's call sites and WebManager the rest) and the rest the new strings. 2026-10-14 fix/web-fewer-writes: override 2n (webserver_fewer_writes.patch), HTTPServer collects a response's small writes in a buffer of one TCP segment and sends them together (#189). Every image grows: release +328 B (1,022,468, .bin 1,034,500), test +328 (1,026,948, .bin 1,038,980, 1,163 B of OTA headroom left), test_https +328 (1,031,956), asserts +328 (1,024,764, .bin 1,036,796), alpha +328 (976,012, the .bin the same), air +328 (1,019,344) with its .bin +4,096 (1,035,340), pico2_w_release +296 (998,600, .bin 998,624), pico_wire +328 (591,548, .bin 603,584), pico_wiretft +328 (773,700, .bin 785,736). The Air's .bin moves in 4 KB steps, as the alpha's does: their .rodata starts on a 4 KB boundary because it holds BTstack's __bluetooth_tlv (8 KB, sector-aligned), and the Air's .text ended 196 B short of the next step. Against main's release the sized symbols grow 362 B and the image 328: _outAppend( ) 140, the constructor's malloc( ) 56, _outBodySent( ) 32, _outFlush( ) 22, and the flushes and resets in send( ), sendContent( ), _prepareHeader( ), _handleRequest( ), _finalizeResponse( ) and the destructor the rest. The 1,460 B buffer is heap, allocated with the server. As an array among the class's members instead of behind a pointer at its end, it pushed every member after it past the Cortex-M0+'s short load/store offsets: release +1,152 B, against +352 B for the same code behind the pointer. The bench ran the release of the branch's first version (sha256 b1a7e7da), which had two more stores, _outLen = 0 in _shortWrite( ), which nothing reaches, and in stopClient( ), which runs but changes nothing: what it would drop goes to a client already closed, whose writes return 0. The final normalized disassembly differs from the bench image's in _shortWrite( ) and dropAbortedStream( ) only, 16 B. 2026-10-15 deps/pico-enc28j60-1.3.1: the ENC28J60 library moves from fb16ec1 to the commit of v1.3.1 (4984e2f), whose lwIP interface keeps up to eight frames in the chip and uploads one while the one before is on the wire (the library's 1.3.0; 1.3.1 changed an example and the documentation only). The wired images grow: pico_wire +568 B (592,116, .bin 604,152), pico_wiretft +624 (774,324, .bin 786,360), RAM +56 on both (the queue's state in the driver object). Nearly all of it is the new send path (_sendQueued( ) 228, _sendPoll( ), _sendFinish( ) and enc28j60_frame_send_finish_blk( ) 100 each, _sendStartNext( ) 80, against sendFrame( ) -144 and enc28j60_frame_send_blk( ) -90); the /license page, whose third-party note names the new commit, 12. The seven radio images are byte for byte main's (192e1ad), compared by sha256. 2026-10-15 fix/wire-tft-flash-render-274: the WiFiServerSecure accept override (wifiserversecure_drop_plaintext.patch, patch.sh 2b-ter) drops a non-TLS client on the TLS port before the BearSSL handshake, decided from its first byte (#274). It is compiled wherever SIMUT_WEB_HTTPS is set, so the five HTTPS images grow: pico_wiretft +56 B (774,380, .bin 786,416), release +48 (1,022,516, .bin 1,034,548), test_https +48 (1,032,004), asserts +56 (1,024,820, .bin 1,036,852), pico2_w_release +24 (998,624, .bin 998,648). pico_wire, pico_w_test, alpha and air compile no HTTPS server, so the patched accept path is not in them and they keep main's sizes (d5f7063). 2026-10-15 perf/tft-spi-clk-peri: setup( ) clocks clk_peri from clk_sys on the six images that drive the panel and SIMUT_TFT_SPI_HZ is 66.5 MHz; /api/status gains sys.clk_peri and sys.tft_hz through macros, so the images without the panel compile the very tokens they did. Flash used (.bin): release +216 (1,022,732, .bin 1,034,764), test +208 (1,027,156, .bin 1,039,188), test_https +208 (1,032,212), asserts +208 (1,025,028, .bin 1,037,060), pico2_w_release +224 (998,848, .bin 998,872), wiretft +208 (774,588, .bin 786,624). pico_wire, alpha and air are byte for byte main's (80c3c88), compared by sha256. 2026-10-14 feat/ui-drivers (S2b-3): pico_wiretft_angulo joins (the Ângulo UI driver's bench image: pico_wiretft with `ui = \"angulo\"`, the touch panel drawn by src/ui/angulo/, transitional and not approved for products): 819,528 B and a .bin of 831,560 B (RAM 116,292 B), +45,828 B and +45,824 B over pico_wiretft (773,700 B, .bin 785,736, RAM 109,784 B), budget = measured. Built against a PlatformIO core that carries overrides 2m and 2n, the private copy that reproduces CI to the byte: in the same session pico_wiretft came out at 773,700 B (.bin 785,736) and pico_w_release at 1,022,468 B (.bin 1,034,500), equal to their records. The image is not exempt from the OTA ceiling: it takes the bench OTA (ota_trust_bench), and signed it is 831,801 B, 208,583 B under 1,040,384; its row is in the OTA headroom table of PLANO_STABLE.md section 3.1. The nine others keep their records. 2026-10-15 feat/ui-drivers-s5 (S5): the Painel group's touches become core actions (src/DisplayManager_Actions.h). pico_w_release, pico_w_test and pico_w_test_https grow 8 B at C3 and pico_w_asserts 8 B at C6 (each of the two steps adds two instructions to handleTouch in all four: 1,022,404, 1,026,884, 1,031,892 and 1,024,708 B used; .bin 1,034,436, 1,038,916 and 1,036,740 B), measured on 60eabd6 (the C8 builds) and again on 33836c2 (the C9 builds), the same sizes and the same sha256 for the nine product images, built against the private core that carries overrides 2m and 2n; pico2_w_release, pico_wiretft, pico_w_air, pico_w_alpha and pico_wire are the size they had at the end of S4. pico_wiretft_angulo 815,424 -> 816,264 B used (+840 B: C2 -32, C3 -88, C4 +136, C5 +328, C6 +424, C7 +72, C8 0), .bin 828,296 B. The review of S5 (C10c): `uiBootSkipHit` asks which frame is on the glass before it reads the frame's targets (`s_modo == MODE_DASHBOARD`, one comparison more), so another screen's target of type 1 cannot read as the boot's skip button. Only pico_wiretft_angulo compiles that file: 816,264 -> 816,280 B used (+16: `uiBootSkipHit` 29 -> 34 instructions, and one padding instruction in `_fini`, the only two functions that differ in the normalized disassembly of the image), .bin 828,296 -> 828,312 B, budget = measured; its row in the OTA headroom table of PLANO_STABLE.md moves with it (211,847 -> 211,831 B of headroom). Measured on a1fda61 plus this change, built against the same private core. 2026-10-15 feat/ui-angulo-visual (the panel's look: its own colour tokens, smoothed corners, the 9 pt text floor), merged over S5 at 2c89d12: pico_wiretft_angulo 816,280 -> 816,104 B used (.bin 828,312 -> 828,136), -176 B: the 512 B the smoothed corners took out on their own (8411d14, measured on the end of S4, which recorded only the numbers; Adafruit GFX's arc helpers are no longer linked) less what the 9 pt layouts added. Then the delivery's SANS7 and SANSB7 leave the kit, which nothing draws since the floor (4,939 B of glyph tables and bitmaps): 811,064 B used, .bin 823,096 B (-5,040 B), budget = measured; its row in the OTA headroom table of PLANO_STABLE.md moves to 217,047 B. The nine product images do not compile src/ui/angulo/ and keep their records. Built against the private core that carries overrides 2m and 2n. 2026-10-15 feat/ui-drivers-s6 (S6, the Base and Menus groups' touches as core actions): the nine product images are byte-equal to the end of S5 at every step and keep their records. pico_wiretft_angulo 811,064 -> 811,232 B used (+168 B: C1 7ee79e9 0, C2 a872d93 +128, C3 77e89aa -24, C4 390fc6f 0, C5 05da4f0 0, C6 80ecdb0 +64, C7 5da8888 0, C8 c078ef4 0), .bin 823,096 -> 823,264 B, budget = measured; its row in the OTA headroom table of PLANO_STABLE.md moves to 216,879 B. C3's -24 B carries a deviation: once actUserDelConfirm( ) added a call site in uaToqueBase( ), GCC stopped inlining pushUiEvent( ) into the Angulo's uiPass( ) (714 -> 674 instructions, one more bl; the out-of-line pushUiEvent( ) is unchanged and only this transitional image is affected; stages/s6_c3/accept_main_session.txt, UI_DRIVERS.md section 2.9). C7 and C8 change no size: C7 moves utf8FoldNext( ) and its three switch tables to an earlier place in the image (no instruction changes; stages/s6_c7/constructor_proof.txt), and C8's src/display/ListPager.h is included by the Angulo's glue only. Built against the private core that carries overrides 2m and 2n. 2026-10-15 feat/ui-drivers-s7 (S7, the Gráficos group's touches as core actions): the nine product images keep their sizes, budgets and OTA rows; C4 64dc6c6 changed the bytes of five of them, in handleTouch( ) only and at the same instruction count (pico_w_release, pico_w_test, pico_w_test_https, pico_w_asserts and pico_wiretft; their hashes are in stages/s7_c4/sha256.txt), and the other four are byte for byte what S6 left. pico_wiretft_angulo 811,232 -> 811,304 B used (+72 B: C1 7fb7d18 0, C2 73ecbd9 0, C3 43960a7 +80, C4 64dc6c6 -8, C5 af96f64 0, C6 6102e9f 0, C7 5e27b04 0), .bin 823,264 -> 823,336 B, budget = measured; its row in the OTA headroom table of PLANO_STABLE.md moves to 216,807 B. C3: handleTouch( ) 4,581 -> 4,603 instructions and uaToqueGraficos( ) 322 -> 346, with drawGraphHeaderBar( ) out of the image; C4: 4,603 -> 4,609 and 346 -> 343, with drawGraphBusyHint( ) out. C6 and C7 change no size: C6 permutes 28 linker veneers among themselves (231 bytes of the .bin, every one a call to a veneer or the veneer's slot; stages/s7_c6/layout_proof.txt) and C7's uaToqueGraficos( ) is two instructions shorter (343 -> 341) inside an image that keeps its size. Built against the private core that carries overrides 2m and 2n.2026-10-15 fix/ui-angulo-aviso: the status bar's badge finds its room between the brand and the icons (src/ui/angulo/telas/Telas.cpp), and the view model carries the notice's kind, seconds and name instead of a composed string; only pico_wiretft_angulo changes, 811,232 -> 811,656 B (+424 B: the width arithmetic, the cut of the name and the words moved from the glue into desenharStatus( )), .bin 823,264 -> 823,688. The nine product images do not compile src/ui/angulo/ (git diff --stat touches only src/ui/angulo/, tools/ and docs/). Built against the private core that carries overrides 2m and 2n; not run on a board. Then the web notice reads \"Web: \", the original top bar's form (its fixed words are 42 px where \" entrou pela web\" took 136, so a 15-character name shows whole beside the cable icon): 811,656 -> 811,640 B (-16 B, the budget comes down with it), .bin 823,688 -> 823,672; the nine product images are still untouched. 2026-10-15, the notice fix merged over S7 (fix/ui-angulo-aviso with be61a2f): pico_wiretft_angulo 811,712 B used, .bin 823,744 B, budget = measured; the nine product images do not compile src/ui/angulo/ and keep S7's records. 2026-10-15, fix/ui-angulo-toque on a6026d5 plus this change (touch targets that answer a finger): pico_wiretft_angulo 811,712 -> 811,928 B (+216 B: the hit test's second pass, the nearest box within 12 px, and the wider rectangles of the title bar's actions and of the limits pager's arrows; not split between the two, built once), .bin 823,744 -> 823,960 B, budget = measured; the nine product images do not compile src/ui/angulo/ (git diff --stat touches only src/ui/angulo/, tools/ and docs/) and keep their records. Built against the private core that carries overrides 2m and 2n, no patch.sh run; not run on a board. 2026-10-15, fix/ui-angulo-historico (a6026d5 plus this change): pico_wiretft_angulo 811,736 B used (+24 B), .bin 823,768 B (+24), budget = measured: graphAnchorStep( ) and graphOpenClearsAnchor( ) in Core 0 under SIMUT_UI_ANGULO; the nine product images do not compile that code (the `#if` is false in them) and keep S7's records. 2026-10-15, the two merged on fix/ui-angulo-navegacao: pico_wiretft_angulo 811,952 B used, .bin 823,984 B, budget = measured (the sum of both, +216 and +24). 2026-10-15, fix/ui-angulo-navegacao (the inert targets, the central band, the page cycle, the month arrows; 2aa380a plus this change, built once): pico_wiretft_angulo 811,952 -> 812,352 B used (+400 B), .bin 823,984 -> 824,384, budget = measured; the targets the screens list for their disabled controls, the band and its page cycle in the glue, and alvoDoToque( ); the control recorder (UA_CONTROLE( )) is ( (void)0 ) on the device. The nine product images do not compile src/ui/angulo/ (git diff --stat touches only src/ui/angulo/, tools/ and docs/) and keep their records. Built against the private core that carries overrides 2m and 2n, no patch.sh run; not run on a board. 2026-10-15, fix/ui-angulo-navegacao, the review's fixes (the host check's pixel rule and the recorder's test, the title as an inert target, the detail's page rule; 67c6521 plus this change, built once): pico_wiretft_angulo 812,352 -> 812,544 B used (+192 B: one more inert target on every screen with a title bar, and the glue's page rule), .bin 824,384 -> 824,576, budget = measured; the host-only code (the control recorder, the drop counter) is ( (void)0 ) or absent on the device. The nine product images do not compile src/ui/angulo/ (git diff --stat touches only src/ui/angulo/, tools/ and docs/) and keep their records. Built against the private core that carries overrides 2m and 2n, no patch.sh run; not run on a board. 2026-10-15, fix/ui-angulo-historico, second version (4122282 plus this change: the live base replaces the half-step bound): pico_wiretft_angulo 811,792 B used (+56 B on 4122282, +80 B on a6026d5), .bin 823,824 B (+56), budget = measured: graphAnchorStep( ) with the base, graphAnchorDay( ) and graphAnchorOpen( ) in Core 0 under SIMUT_UI_ANGULO; the nine product images do not compile that code and keep S7's records. 2026-10-15, fix/ui-angulo-historico (d2bf3cd, 456a1ad) merged over fix/ui-angulo-navegacao 91df450: pico_wiretft_angulo 812,600 B used, .bin 824,632 B, budget = measured. 2026-10-15, feat/ui-drivers-s8a (S8a, C1 to C8 over 9bef346, no patch.sh run, not run on a board): the nine product images moved in handleTouch( ) at C2 to C5 and kept their sizes but for C5, which took 8 B off six of them (pico_w_release 1,022,396 B used, .bin 1,034,428; pico_w_test 1,026,876, .bin 1,038,908; pico_w_test_https 1,031,884; pico_w_asserts 1,024,700, .bin 1,036,732; pico2_w_release 998,528, .bin 998,552; pico_wiretft 773,628, .bin 785,664); pico_wiretft_angulo 812,600 -> 813,048 B used (+448: C2 -48, C3 +144, C4 +320, C5 -8, C6 +40, C7 and C8 0), .bin 824,632 -> 825,080, budget = measured. Each action is compiled twice in that image (inlined into handleTouch( ) and into its own glue) until S10b. 2026-10-15, fix/ui-angulo-margem (3ba3ea5 plus this change, no patch.sh run, not run on a board): the panel's 4 px margin is black in both themes (ua::margemPreta( ) after each strip's draw in uaPintar( ), and the glass cleared to black on a full repaint): pico_wiretft_angulo 813,048 -> 813,152 B used (+104 B), .bin 825,080 -> 825,184, budget = measured; its row in the OTA headroom table of PLANO_STABLE.md moves to 214,959 B. The nine product images do not compile src/ui/angulo/ and keep their records (git diff --stat touches only src/ui/angulo/, tools/ and docs/). 2026-10-15 perf/ui-angulo-pintura (p1, a frame is drawn once and replayed): pico_wiretft_angulo 813,152 -> 814,496 B (+1,344), .bin 825,184 -> 826528; RAM 116,300 -> 131,156 B (+14,856, the record: 1024 calls of 14 B and 512 B of text); the nine product images do not compile src/ui/angulo/. 2026-10-15 perf/ui-angulo-pintura (p2, the strip hash by words): pico_wiretft_angulo 814,496 -> 814,664 B (+168), .bin 826,528 -> 826,696; RAM 131,156 -> 131,156 B (+0). 2026-10-15 perf/ui-angulo-pintura (p3, the corner and disc coverage tables): pico_wiretft_angulo 814,664 -> 816,792 B (+2,128), .bin 826,696 -> 828,824; RAM 131,156 -> 131,156 B (+0). 2026-10-15 perf/ui-angulo-pintura (p4, glyphs straight into the strip): pico_wiretft_angulo 816,792 -> 817,208 B (+416), .bin 828,824 -> 829,240; RAM 131,156 -> 131,156 B (+0). 2026-10-15 perf/ui-angulo-pintura (p5, fills by words): pico_wiretft_angulo 817,208 -> 817,256 B (+48), .bin 829,240 -> 829,288; RAM 131,156 -> 131,156 B (+0). 2026-10-15 perf/ui-angulo-pintura (p6, no black frame): pico_wiretft_angulo 817,256 -> 817,016 B (-240), .bin 829,288 -> 829,048; RAM 131,156 -> 131,156 B (+0). 2026-10-15 perf/ui-angulo-pintura (p7, the record in the canvas's spare rows, the chart's plot as three repeatable calls): pico_wiretft_angulo 817,016 -> 817,448 B (+432), .bin 829,048 -> 829,480; RAM 131,156 -> 116,348 B (-14,808). 2026-10-15 perf/ui-angulo-pintura (p8, the strips pushed by DMA while the next is composed, in src/ui/angulo/DisplayManager_UiA_Empurrar.cpp; the part of a strip on the glass at an alignment offset too): pico_wiretft_angulo 817,448 -> 817,424 B (-24), .bin 829,480 -> 829,456; RAM 116,348 -> 116,376 B (+28). 2026-10-16 perf/ui-angulo-pintura (p9, the review's fixes: the record's list returned after a paint, the in-flight flag set before the bus opens): pico_wiretft_angulo 817,424 -> 817,432 B (+8), .bin 829,456 -> 829,464; RAM 116,376 -> 116,376 B (+0). 2026-10-16 perf/ui-angulo-pintura (p10, line( ), fillRect( ) and rect( ) drop what lies outside the strip's window): pico_wiretft_angulo 817,432 -> 817,504 B (+72), .bin 829,464 -> 829,536; RAM 116,376 -> 116,376 B (+0). 2026-10-16 perf/ui-angulo-pintura (p11a, the plot spares the buckets a strip cannot show): pico_wiretft_angulo 817,504 -> 817,920 B (+416), .bin 829,536 -> 829,952; RAM 116,376 -> 116,376 B (+0). 2026-10-16 perf/ui-angulo-pintura (p11b, rounded boxes: runs as spans, in-place blend, body in three pieces): pico_wiretft_angulo 817,920 -> 818,200 B (+280), .bin 829,952 -> 830,232; RAM 116,376 -> 116,376 B (+0). 2026-10-16 perf/ui-angulo-pintura (p12a, the strip's fill and hash run from RAM): pico_wiretft_angulo 818,200 -> 818,216 B (+16), .bin 830,232 -> 830,248; RAM 116,376 -> 116,744 B (+368). 2026-10-16 sync/ui-drivers-main (main 2d6650b merged into feat/ui-drivers 1a913a4, so that the branch carries #275 PicoENC28J60 1.3.1, #276, #277 and #278): every image rebuilt on the merged tree against the default PlatformIO core (~/.platformio, framework 1.50601.0), which carries every override of patch.sh, 2b-ter (the WiFiServerSecure accept patch of #277) included; pico_w_release was built a second time at the end and came out equal. The first measurement of this tree, on the private core of the earlier entries, which lacks 2b-ter, is replaced: pico_w_test, pico_w_alpha, pico_w_air and pico_wire are byte for byte the same on both cores, and the HTTPS images grew by what main measured for that patch (pico_wiretft +56 B, release +48, test_https +48, asserts +56, pico2_w_release +24), pico_wiretft_angulo by +48. Flash used (.bin) against the branch's records: release 1,022,396 -> 1,022,660 (+264, 1,034,692), test +216 (1,027,092, 1,039,124), test_https +256 (1,032,140), asserts +264 (1,024,964, 1,036,996), pico2_w_release +248 (998,776, 998,800), pico_wire +568 (592,116, 604,152; the 1.3.1 send path, as main measured), pico_wiretft +896 (774,524, 786,560), pico_wiretft_angulo +904 (819,120, 831,152; RAM 116,744 -> 116,812 B). pico_w_alpha and pico_w_air are byte for byte the branch's previous images (sha256 against stages/s8a_c5): the clock code is compiled out where there is no panel. 2026-10-21 sync2/ui-drivers-main: main 90e82d1 (#279, #280) merged into d9cdda4. #279 touches no firmware; #280 changes how an injected touch ends, in handleTouch( ), loopCore1( ) and mapTouchPoint( ). All ten images rebuilt from scratch with the default core, against the same ten of d9cdda4 built the same way. Flash used: release +48 B (used 1,022,700, .bin 1,034,732), test +56 B (used 1,027,132, .bin 1,039,164), test_https +48 B (used 1,032,180, .bin 1,044,212), asserts +56 B (used 1,025,004, .bin 1,037,036), pico2_w_release +16 B (used 998,784, .bin 998,808), wiretft +56 B (used 774,572, .bin 786,608); alpha, air, wire, wiretft_angulo byte for byte the same. On main the same change was release +40, test +40, test_https +40, asserts +40, pico2_w_release +24 and wiretft +48; the branch compiles handleTouch( ) and loopCore1( ) differently around it. In the normalized disassembly (tools/elf_funcoes.py) each image that changed differs from d9cdda4 in loopCore1( ), mapTouchPoint( ) and handleTouch( ) only, plus _fini( ) where its trailing alignment halfword came or went, and on the Pico 2 W the binary-info header word that holds an address. 2026-10-21 S9 C2 (feat/ui-drivers-s9): the PIN keypad's touch becomes core actions, compiled into the Angulo's glue as well as into the original's handleTouchPanelV24( ), which the Angulo image links until S10b: pico_wiretft_angulo 821,608 B (+600 B) and a .bin of 833,640 B; the nine product images byte for byte unchanged. 2026-10-21 S9 C3 (feat/ui-drivers-s9): the sensor menu's permission table moves into its core action, so its closure, an IPA-SRA clone now, builds the table itself and handleTouchPanelV24( ) no longer does: the six TFT product images 8 B smaller (release, test, test_https, asserts, pico2_w_release, pico_wiretft; budgets lowered with them), pico_wiretft_angulo 16 B larger (821,624 B, .bin 833,656 B); same calls in the same order and the same conditional branches in every changed function (tools_s9/calls_check.py). 2026-10-21 S9 C4 (feat/ui-drivers-s9): the account editor's permission table moves into its core action, which the handler's inlined closure now builds in the row's own path, so handleTouchPanelV24( )'s frame is 8 B smaller: five TFT product images 8 B smaller (release, test, test_https, asserts, pico_wiretft; budgets lowered with them), pico2_w_release the same size with different bytes, pico_wiretft_angulo 8 B larger (821,632 B, .bin 833,664 B); same calls in the same order and the same conditional branches in every changed product function (tools_s9/calls_check.py, which counts the RP2350's cbz / cbnz as the beq / bne they stand for since this step). 2026-10-21 feat/ui-drivers-s9, S9 C6 (the password keyboard's typing as two core actions): the nine product images are byte-equal to C5's and C4's; only pico_wiretft_angulo changes, 8 B smaller (its digitar( ) lambda calls actPwdAppend( ) instead of storing twice): measured 821624 B and .bin 833656 B, the numbers of C3, so its budget goes down with it and its OTA row too. 2026-10-18 fix/safe-reboot-writes-pending-log: safeReboot( ) writes the log records it holds in RAM (flushPendingIfAny( ) inside the pre-reboot hook's guard, in safeReboot( ) and in the RP2350's safeRebootFlashUpdate( )). Four images grow 8 B, the call: pico_w_test (1,027,204, .bin 1,039,236), pico_w_asserts (1,025,076, .bin 1,037,108), pico2_w_release (998,880, .bin 998,904) and pico_wire (592,124, .bin 604,160). pico_w_release, pico_w_test_https, pico_w_alpha, pico_w_air and pico_wiretft come out the size of their records, flash used and .bin, their alignment absorbing it; RAM unchanged. 2026-10-21 fix/radio-answers-by-return-code (#266, #282): a reconnect scan that started quickly and never finished asks the chip one read (radioRead( ), an antenna-setting ioctl) and goes by the driver's return code, a quick start no longer ends a silent streak, and the minute's RSSI sample and getRssi( ) read cyw43_wifi_get_rssi( )'s return code (readRssi( )) instead of the variable WiFi.RSSI( ) leaves unwritten. On main 7e3f29a6 (#281 in), the seven radio images grow: release +160 B (1,022,932, .bin 1,034,964), test +176 (1,027,380, .bin 1,039,412, 731 B of OTA headroom left), test_https +176 (1,032,428), asserts +160 (1,025,236, .bin 1,037,268), alpha +168 (976,180) and air +168 (1,019,512), both .bin the same, pico2_w_release +184 (999,064, .bin 999,088). The two wired images build no radio unit and are byte for byte main's (pico_wire sha256 f464769b, pico_wiretft e781ea41); RAM unchanged on all nine. 2026-10-21 sync3/ui-drivers-main: feat/ui-drivers 154ff60c takes main 2f4bc3d3 (#281, #282; the merge commit that carries this note). All ten images rebuilt from scratch on the merged tree, against ~/.platformio, and 154ff60c rebuilt the same way for the other side (flash used and .bin, before -> after): pico_w_release 1,022,684 -> 1,022,852 B used (+168), .bin 1,034,716 -> 1,034,884 (+168); pico_w_test 1,027,116 -> 1,027,292 B used (+176), .bin 1,039,148 -> 1,039,324 (+176); pico_w_test_https 1,032,164 -> 1,032,348 B used (+184), .bin 1,044,196 -> 1,044,380 (+184); pico_w_asserts 1,024,988 -> 1,025,148 B used (+160), .bin 1,037,020 -> 1,037,180 (+160); pico_w_alpha 976,012 -> 976,180 B used (+168), .bin 990,380 -> 990,380 (+0); pico_w_air 1,019,344 -> 1,019,512 B used (+168), .bin 1,035,340 -> 1,035,340 (+0); pico2_w_release 998,776 -> 998,968 B used (+192), .bin 998,800 -> 998,992 (+192); pico_wire 592,116 -> 592,124 B used (+8), .bin 604,152 -> 604,160 (+8); pico_wiretft 774,556 -> 774,564 B used (+8), .bin 786,592 -> 786,600 (+8); pico_wiretft_angulo 821,624 -> 821,632 B used (+8), .bin 833,656 -> 833,664 (+8). Every record above is the merged tree's measurement, not the sum of the two sides' text; the two sentences just above this one (2026-10-18 and 2026-10-21) are main's own, with main's sizes, on main's trees. Host only: nothing of this merge has run on a board. 2026-10-21 feat/ui-drivers-s10a C2: the limits editor's step in one action for both drivers (form Fb), the one step of S10a that changes the products; the three LCD images are byte for byte C0's and the seven TFT images were built here by tools_s10a/verify_step.sh: pico_w_release 1,022,852 -> 1,022,940 B used (+88), .bin 1,034,884 -> 1,034,972 (+88); pico_w_test 1,027,292 -> 1,027,388 B used (+96), .bin 1,039,324 -> 1,039,420 (+96); pico_w_test_https 1,032,348 -> 1,032,436 B used (+88), .bin 1,044,380 -> 1,044,468 (+88); pico_w_asserts 1,025,148 -> 1,025,244 B used (+96), .bin 1,037,180 -> 1,037,276 (+96); pico2_w_release 998,968 -> 998,960 B used (-8), .bin 998,992 -> 998,984 (-8); pico_wiretft 774,564 -> 774,660 B used (+96), .bin 786,600 -> 786,696 (+96); pico_wiretft_angulo 821,632 -> 821,688 B used (+56), .bin 833,664 -> 833,720 (+56). Host only: nothing of this change has run on a board. 2026-10-21 fix/language-pack-log-records (#283): the language pack's records get their own tag (TAG_I18N, 13) and codes (620-623), the console text of the four carries only its detail, and the tags passed to the LOG_ macros are checked at compile time. On main 2f4bc3d3, the seven images that build the pack loader shrink: release -24 B (1,022,908, .bin 1,034,940), test -32 (1,027,348, .bin 1,039,380, 763 B of OTA headroom left), test_https -24 (1,032,404), asserts -24 (1,025,212, .bin 1,037,244), air -32 (1,019,480, the .bin the same), pico2_w_release -48 (999,016, .bin 999,040), wiretft -24 (774,612, .bin 786,648); their budgets come down. On release: findAndLoadLangFile( ) -92 (its two String concatenations), translateCodeEn( ) +52 (the four new cases), tagStringToId( ) and loadLangFile( ) +4 each, .rodata +8 (the four names against the shorter console texts); the name table moved from tagIdToString( )'s function-local static to LOG_TAG_NAMES, 64 B either way. The two images that do not build the loader grow, raising their budgets: pico_wire +184 (592,308, .bin 604,344: translateCodeEn( ) +52 and tagStringToId( ) +4 in .text, the four names +128 in .rodata) and alpha +56 (976,236, the .bin the same: .text +56, its .rodata absorbing the names in the padding before BTstack's sector-aligned TLV). RAM unchanged on all nine. The check itself costs nothing: the same tree with no static_assert, tagStringToId( ) inline and the old function-local table builds byte-identical images (release sha256 e971dfe8, air e5935de2, wiretft a01a20b2). 2026-10-21 sync4/ui-drivers-main: feat/ui-drivers edc95d6d takes main 25e169df (#284, the language pack's log records; the merge commit that carries this note). All ten images rebuilt from scratch on the merged tree, against ~/.platformio, and edc95d6d rebuilt the same way for the other side (flash used and .bin, before -> after): pico_w_release 1,022,940 -> 1,022,916 B used (-24), .bin 1,034,972 -> 1,034,948 (-24); pico_w_test 1,027,388 -> 1,027,356 B used (-32), .bin 1,039,420 -> 1,039,388 (-32); pico_w_test_https 1,032,436 -> 1,032,412 B used (-24), .bin 1,044,468 -> 1,044,444 (-24); pico_w_asserts 1,025,244 -> 1,025,220 B used (-24), .bin 1,037,276 -> 1,037,252 (-24); pico_w_alpha 976,180 -> 976,236 B used (+56), .bin 990,380 -> 990,380 (+0); pico_w_air 1,019,512 -> 1,019,480 B used (-32), .bin 1,035,340 -> 1,035,340 (+0); pico2_w_release 998,968 -> 998,920 B used (-48), .bin 998,992 -> 998,944 (-48); pico_wire 592,124 -> 592,308 B used (+184), .bin 604,160 -> 604,344 (+184); pico_wiretft 774,660 -> 774,636 B used (-24), .bin 786,696 -> 786,672 (-24); pico_wiretft_angulo 821,432 -> 821,400 B used (-32), .bin 833,464 -> 833,432 (-32). Every record above is the merged tree's measurement, not the sum of the two sides' text; the sentence just above this one is main's own, with main's sizes, on main's trees. Host only: nothing of this merge has run on a board. 2026-10-21 s10b/c0b: pico2_w_angulo joins, the Angulo on the Pico 2 W (pico2_w_release with ui = angulo, publish = false; like its parent it does not trust the bench root). The ten images and the new one rebuilt from scratch on feat/ui-drivers-s10b 50828bdf plus this change, against ~/.platformio: pico2_w_angulo 1,041,276 B used, .bin 1,041,296, RAM 132,564 B (budget = used; the slot ceiling 1,568,768 B as on pico2_w_release); against pico2_w_release on the same tree (998,920 B used, .bin 998,944) it is +42,356 B used and +42,352 B of .bin. The ten others are byte for byte what sync4 and C0 recorded (the nine sha256 and the Wire Angulo 44e1220f), so no other record moves. Host only: nothing of this change has run on a board. 2026-10-21 s10b/c2: pico_wiretft_angulo 821,400 -> 821,440 B used (+40), .bin 833,432 -> 833,472 (+40); pico2_w_angulo 1,041,276 -> 1,041,308 B used (+32), .bin 1,041,296 -> 1,041,328 (+32) (built from scratch with tools/build9.sh after the update screen's case called actOtaDismiss( ) itself and uaToque( ) answered true on every path). 2026-10-21 s10b/c3: pico_wiretft_angulo 821,440 -> 808,848 B used (-12,592), .bin 833,472 -> 820,880 (-12,592); pico2_w_angulo 1,041,308 -> 1,029,876 B used (-11,432), .bin 1,041,328 -> 1,029,896 (-11,432) (built from scratch with tools/build9.sh after the original's touch chain moved out of the Angulo's link (about 12.5 kB of the chain and handleTouchPanelV24( ) and what only they reached)). 2026-10-21 s10b/c4a: pico_wiretft_angulo 808,848 -> 808,728 B used (-120), .bin 820,880 -> 820,760 (-120); pico2_w_angulo 1,029,876 -> 1,029,748 B used (-128), .bin 1,029,896 -> 1,029,768 (-128) (built from scratch with tools/build9.sh after Themes.cpp left the Angulo's link and the Angulo defined the theme catalogue in ui/angulo/Temas.cpp (its three rows; the original's palettes, the .thm parser and its 800 B of custom themes leave)). 2026-10-21 s10b/c4b: pico_wiretft_angulo 808,728 -> 808,904 B used (+176), .bin 820,760 -> 820,936 (+176); pico2_w_angulo 1,029,748 -> 1,029,948 B used (+200), .bin 1,029,768 -> 1,029,968 (+200) (the .thm upload is refused at UPLOAD_FILE_START where SIMUT_UI_CUSTOM_THEMES is 0 (the Angulo images): WebManager::handleUploadData( ) and the new text; the nine compile it out). 2026-10-21 s10b/c5b: pico_wiretft_angulo 808,904 -> 808,896 B used (-8), .bin 820,936 -> 820,928 (-8); pico2_w_angulo 1,029,948 -> 1,029,932 B used (-16), .bin 1,029,968 -> 1,029,952 (-16) (the _uaPintando guards of fastFillRect, fastClearScreen and blitCanvas, the member and its two writes leave: the Angulo image does not link the three primitives; only the constructor, uaPintar and the offset of the member after the removed one change). S10c A'4, 2026-10-22: pico_wiretft_angulo 808,896 -> 808,536 B used (-360), .bin 820,928 -> 820,568 (-360); pico2_w_angulo 1,029,932 -> 1,029,556 B used (-376), .bin 1,029,952 -> 1,029,576 (-376) (a .thm upload is stored, not refused: C4b's refusal block and the two theme rescans leave the Angulo images, three strings go (.thm, Custom themes rescanned, the refusal text); only WebManager::handleUploadData and WebManager::handleDelete change in code, the rest is layout). 2026-10-22 s10c/a2: pico_wiretft_angulo 808,536 -> 809,408 B used (+872), .bin 820,568 -> 821,440 (+872); pico2_w_angulo 1,029,556 -> 1,030,524 B used (+968), .bin 1,029,576 -> 1,030,544 (+968) (A'2 puts the palette's 24 generated tables and the mode and accent functions into the two Angulo images). 2026-10-22 s10c/a3: pico_wiretft_angulo 809,408 -> 811,664 B used (+2,256), .bin 821,440 -> 823,696 (+2,256); pico2_w_angulo 1,030,524 -> 1,032,796 B used (+2,272), .bin 1,030,544 -> 1,032,816 (+2,272) (A'3's Tema screen (segments, swatches, button), the preview and the catalogue of 18 enter the two Angulo images). 2026-10-22 s10c/a3: pico_wiretft_angulo 811,664 -> 811,664 B used (+0), .bin 823,696 -> 823,696 (+0); pico2_w_angulo 1,032,796 -> 1,032,796 B used (+0), .bin 1,032,816 -> 1,032,816 (+0) (A'3's Tema screen (segments, swatches, button), the preview and the catalogue of 18 enter the two Angulo images). 2026-10-22 s10c/a2b: pico_wiretft_angulo 811,664 -> 811,300 B used (-364), .bin 823,696 -> 823,336 (-360); pico2_w_angulo 1,032,796 -> 1,032,400 B used (-396), .bin 1,032,816 -> 1,032,424 (-392) (the Angulo's palette is built by the compiler (constexpr tables from the generator) and the static initialiser is gone). 2026-10-22 s10c/b2b: pico_wiretft_angulo 811,300 -> 812,244 B used (+944), .bin 823,336 -> 824,280 (+944); pico2_w_angulo 1,032,400 -> 1,033,304 B used (+904), .bin 1,032,424 -> 1,033,328 (+904) (the Angulo loads its texts from the language pack at boot: Textos.cpp joins the two images, the hook, the scanner's @ANGULO entry, two words in ActiveLang). 2026-10-22 S10c B3 (the Painel and Base screens look their texts up in the language pack): pico_wiretft_angulo 812,244 -> 812,972 B used (+728), .bin 824,280 -> 825,008 (+728); pico2_w_angulo 1,033,304 -> 1,033,976 B used (+672), .bin 1,033,328 -> 1,034,000 (+672) (the English literals and the lookups replace the Portuguese literals; the nine products are byte-identical). 2026-10-22 feat/ui-drivers-s10c B3-b: pico_wiretft_angulo 812,972 -> 812,924 B used (-48), .bin 825,008 -> 824,960 (-48); pico2_w_angulo 1,033,976 -> 1,033,920 B used (-56), .bin 1,034,000 -> 1,033,944 (-56) (the kit's back action is one function, acaoVoltar( ), which the three screens that rebuilt it with the literal now call; static RAM unchanged, 115,912 and 131,572 B). 2026-10-22 feat/ui-drivers-s10c B3-c: pico_wiretft_angulo 812,924 -> 812,940 B used (+16), .bin 824,960 -> 824,976 (+16); pico2_w_angulo 1,033,920 -> 1,033,936 B used (+16), .bin 1,033,944 -> 1,033,960 (+16) (the pager's count is one kit function, contagemDoPaginador( ), which paginador( ) draws and the charts' page name is placed after; static RAM unchanged, 115,912 and 131,572 B). 2026-10-22 feat/ui-drivers-s10c B4a: pico_wiretft_angulo 812,940 -> 813,588 B used (+648), .bin 824,976 -> 825,624 (+648); pico2_w_angulo 1,033,936 -> 1,034,544 B used (+608), .bin 1,033,960 -> 1,034,568 (+608) (the Menus screens' texts are UA_TXT( ) lookups, the English in the source and the Portuguese or the Spanish in the packs; static RAM unchanged, 115,912 and 131,572 B). 2026-10-22 feat/ui-drivers-s10c B4b-0: pico_w_release 1,022,916 -> 1,022,900 B used (-16), .bin 1,034,948 -> 1,034,932 (-16); pico_w_test 1,027,356 -> 1,027,348 B used (-8), .bin 1,039,388 -> 1,039,380 (-8); pico_w_asserts 1,025,220 -> 1,025,204 B used (-16), .bin 1,037,252 -> 1,037,236 (-16); pico2_w_release 998,920 -> 998,928 B used (+8), .bin 998,944 -> 998,952 (+8); pico_wire 592,308 -> 592,268 B used (-40), .bin 604,344 -> 604,304 (-40); pico_wiretft 774,636 -> 774,596 B used (-40), .bin 786,672 -> 786,632 (-40); pico_wiretft_angulo 813,588 -> 813,548 B used (-40), .bin 825,624 -> 825,584 (-40); pico2_w_angulo 1,034,544 -> 1,034,552 B used (+8), .bin 1,034,568 -> 1,034,576 (+8) (LANG_FILE_MAX 49,152 -> 65,536 and lang.js's comment of the viewer's cap: the comment re-stamps every web page (the ?v= hash of the raw lang.js), so each page's gzip moves by a few bytes; the normalized pages are identical, and loadLangFile( )'s constant is the only code change). 2026-10-22 feat/ui-drivers-s10c B4b: pico_wiretft_angulo 813,548 -> 814,028 B used (+480), .bin 825,584 -> 826,064 (+480); pico2_w_angulo 1,034,552 -> 1,035,008 B used (+456), .bin 1,034,576 -> 1,035,032 (+456) (the Ajustes screens' texts are UA_TXT( ) lookups, the English in the source and the Portuguese or the Spanish in the packs, and the footer's boxes keep the labels they measure; static RAM unchanged, 115,912 and 131,572 B). 2026-10-22 feat/ui-drivers-s10c B5a: pico_wiretft_angulo 814,028 -> 814,356 B used (+328), .bin 826,064 -> 826,392 (+328); pico2_w_angulo 1,035,008 -> 1,035,392 B used (+384), .bin 1,035,032 -> 1,035,416 (+384) (the Gráficos screens' texts are UA_TXT( ) lookups, the English in the source and the Portuguese or the Spanish in the packs, n1( ) and kb( ) write the language's decimal separator; static RAM unchanged, 115912 and 131572 B). 2026-10-23 feat/ui-drivers-s10c B5a-b: pico_wiretft_angulo 814,356 -> 814,364 B used (+8), .bin 826,392 -> 826,400 (+8); pico2_w_angulo 1,035,392 -> 1,035,400 B used (+8), .bin 1,035,416 -> 1,035,424 (+8) (ua::separadorDecimal( ) takes only ',' or '.' from a pack; static RAM unchanged). 2026-10-23 feat/ui-drivers-s10c B5b: pico_wiretft_angulo 814,364 -> 814,700 B used (+336), .bin 826,400 -> 826,736 (+336); pico2_w_angulo 1,035,400 -> 1,035,592 B used (+192), .bin 1,035,424 -> 1,035,616 (+192) (the Identidade screens' texts are UA_TXT( ) lookups, the English in the source and the Portuguese or the Spanish in the packs, the identity bar and footers keep the labels their boxes were measured from, the Painel's missing-value placeholder takes the language's separator; static RAM unchanged, 115912 and 131572 B). 2026-10-22 fix/alpha-console-event-names: the alpha's two log lookups, logcodeLookup( ) and trlLookup( ) in DisplayManager_Alpha.cpp, answer nullptr where they answered \"\", which LogManager took as the translation. Each became `movs r0, #0; bx lr`, 4 B, where it loaded a literal, 8 B: pico_w_alpha -8 B (976,228, the .bin the same) and pico_wire -8 B (592,300, .bin 604,336), the two images that build that file. The seven others are byte for byte main's (25e169d), compared by sha256. RAM unchanged on all nine. 2026-10-22 chore/https-test-image-room: pico_w_test_https serves the license page from LittleFS too (custom_fs_pages ALARMS_PAGE, FILE_PAGE, TEL_PAGE and LICENSE_PAGE), the maintainer's option (a) for feat/ui-drivers' S10b-2, which grows this image about 160 B where its FLASH region had 44 B left (the last LOAD ended at 0x100FEFD4, readelf -l): -4,256 B (1,028,148, the .bin 1,040,180 and not recorded, ota_exempt), the region's room 4,300 B (the last LOAD ends at 0x100FDF34); its budget comes down. The license page is 4,260 B of gzip in this env. The other eight are byte for byte main's (a8ccf4c3), compared by sha256. 2026-10-23 sync5/ui-drivers-main: feat/ui-drivers ecb527e3 takes main 2075b84e (#285 to #288: the savings floor past the BTstack fill, the console and syslog event names on alpha and wire, the docs of the 22/10 bench, and pico_w_test_https's licence page on LittleFS; the merge commit that carries this note). All eleven images rebuilt from scratch on the merged tree, against ~/.platformio, and ecb527e3 rebuilt the same way for the other side (flash used and .bin, before -> after): pico_w_release 1,022,900 -> 1,022,900 B used (+0), .bin 1,034,932 -> 1,034,932 (+0); pico_w_test 1,027,348 -> 1,027,348 B used (+0), .bin 1,039,380 -> 1,039,380 (+0); pico_w_test_https 1,032,412 -> 1,028,148 B used (-4264), .bin 1,044,444 -> 1,040,180 (-4264); pico_w_asserts 1,025,204 -> 1,025,204 B used (+0), .bin 1,037,236 -> 1,037,236 (+0); pico_w_alpha 976,236 -> 976,228 B used (-8), .bin 990,380 -> 990,380 (+0); pico_w_air 1,019,480 -> 1,019,480 B used (+0), .bin 1,035,340 -> 1,035,340 (+0); pico2_w_release 998,928 -> 998,928 B used (+0), .bin 998,952 -> 998,952 (+0); pico_wire 592,268 -> 592,260 B used (-8), .bin 604,304 -> 604,296 (-8); pico_wiretft 774,596 -> 774,596 B used (+0), .bin 786,632 -> 786,632 (+0); pico_wiretft_angulo 814,716 -> 814,716 B used (+0), .bin 826,752 -> 826,752 (+0); pico2_w_angulo 1,035,608 -> 1,035,608 B used (+0), .bin 1,035,632 -> 1,035,632 (+0). Every record above is the merged tree's measurement, not the sum of the two sides' text; the sentence just above this one is main's own, with main's sizes, on main's trees. Host only: nothing of this merge has run on a board. S10c D5 (2026-10-23): pico_w_release 1,022,900 -> 1,022,908 B used (+8), .bin 1,034,932 -> 1,034,940 (+8); pico_w_test_https 1,028,148 -> 1,028,156 B used (+8), exempt from the OTA ceiling so no .bin recorded; pico2_w_release 998,928 -> 998,952 B used (+24), .bin 998,952 -> 998,976 (+24) (the alarm screen's three draw functions moved to DisplayManager_AlarmUi.cpp, so redrawAlarmFlash( ) reads _lastRenderedState.topSlotIdx again after isSlotAlarming( ) and isSlotErrAlarming( ), which it can no longer see only read (141 -> 143 instructions on the RP2040, 104 -> 110 on the Pico 2 W, where slotAlarmBg( ) also saves this and the slot in r4 and r5 across its calls); host only, nothing ran on a board). S10c D3a: pico_w_release 1,022,908 -> 1,022,916 B used (+8), .bin 1,034,940 -> 1,034,948 (+8); pico_w_test 1,027,348 -> 1,027,364 B used (+16), .bin 1,039,380 -> 1,039,396 (+16); pico_w_test_https 1,028,156 -> 1,028,164 B used (+8), exempt from the OTA ceiling so no .bin recorded; pico_w_asserts 1,025,204 -> 1,025,220 B used (+16), .bin 1,037,236 -> 1,037,252 (+16); pico2_w_release 998,952 -> 998,960 B used (+8), .bin 998,976 -> 998,984 (+8); pico_wiretft 774,596 -> 774,612 B used (+16), .bin 786,632 -> 786,648 (+16); pico_wiretft_angulo 814,716 -> 814,644 B used (-72), .bin 826,752 -> 826,680 (-72); pico2_w_angulo 1,035,608 -> 1,035,528 B used (-80), .bin 1,035,632 -> 1,035,552 (-80) (the seven calibration draw-cache stores moved into resetTouchCalCaches( ) behind UiChange::TOUCH_CAL_ENTER). S10c D3b: pico_wiretft_angulo 814,644 -> 814,636 B used (-8), .bin 826,680 -> 826,672 (-8); pico2_w_angulo 1,035,528 -> 1,035,536 B used (+8), .bin 1,035,552 -> 1,035,560 (+8) (showSettingsDisplayOffset( ) sets the original's offset sentinels behind UiChange::DISPLAY_OFFSET_ENTER). S10c D6a: pico_w_release 1,022,916 -> 1,023,084 B used (+168), .bin 1,034,948 -> 1,035,116 (+168); pico_w_test 1,027,364 -> 1,027,524 B used (+160), .bin 1,039,396 -> 1,039,556 (+160); pico_w_test_https 1,028,164 -> 1,028,324 B used (+160); pico_w_asserts 1,025,220 -> 1,025,380 B used (+160), .bin 1,037,252 -> 1,037,412 (+160); pico2_w_release 998,960 -> 999,072 B used (+112), .bin 998,984 -> 999,096 (+112); pico_wiretft 774,612 -> 774,772 B used (+160), .bin 786,648 -> 786,808 (+160); pico_wiretft_angulo 814,636 -> 814,700 B used (+64), .bin 826,672 -> 826,736 (+64); pico2_w_angulo 1,035,536 -> 1,035,592 B used (+56), .bin 1,035,560 -> 1,035,616 (+56) (the original's user and PIN screens move to DisplayManager_UsersUi.cpp, and GCC inlines PinKb::orderedFace, clampPinPolicy and maxLenFor once their callers are in two units). 2026-10-28 S10c D7: pico2_w_angulo 1,035,592 -> 1,035,992 B used (+400), .bin 1,035,616 -> 1,036,016 (+400) (the setup-network confirmation's frame and touch, three texts, one #if SIMUT_NET_WIFI). 2026-10-24 fix/lang-english-way-back: after a boot in English the Idioma screen lists the pack in /lang without loading it (display/LangChoice.h, listLangFile( ), the choice saved at once and loaded at the next boot): pico_w_release +616 B (1,023,524, the .bin 1,035,556), pico_w_test +624 (1,027,972, the .bin 1,040,004: 139 B of OTA headroom left of 763), pico_w_asserts +608 (1,025,820, the .bin 1,037,852), pico2_w_release +656 (999,672, the .bin 999,696), pico_wiretft +624 (775,236, the .bin 787,272), and pico_w_test_https +624 (1,028,772, the .bin 1,040,804 and not recorded, ota_exempt), which leaves its FLASH region 3,676 B of room (the last LOAD ends at 0x100FE1A4, readelf -l). pico_w_air -16 B (1,019,464, the .bin the same size, 1,035,340): its sha256 differs, because loadLangFile( ) and findAndLoadLangFile( ) are compiled differently (the one rule for a usable pack, and the directory pass the list shares); pico_w_alpha and pico_wire are byte for byte main's (be56633d), compared by sha256. RAM unchanged on all nine: a listed pack is told by the path kept in _activeLang, not by a new variable. 2026-10-28 fix/panel-theme-lang-permission on d2c366b3 (PR #294's head): the guard arm that refuses EVT_APPLY_THEME and EVT_APPLY_LANG without PERM_SYS_CONFIG sits in the dispatch chain's TFT region, so only the six panel images compile it: release +64 B (1,023,588, .bin 1,035,620), test +64 (1,028,036, .bin 1,040,068; OTA room 139 -> 75 B), test_https +72 (1,028,844), asserts +72 (1,025,892, .bin 1,037,924), pico2_w_release +24 (999,696, .bin 999,720), pico_wiretft +64 (775,300, .bin 787,336); pico_w_alpha, pico_w_air and pico_wire are byte for byte d2c366b3's, compared by sha256 (b5d7d079, 4291184e, 3438b89a). RAM unchanged on all nine. 2026-10-28 sync 6 (C1): pico_w_release 1,023,084 -> 1,023,764 B used (+680), .bin 1,035,116 -> 1,035,796 (+680); pico_w_test 1,027,524 -> 1,028,212 B used (+688), .bin 1,039,556 -> 1,040,244 (+688); pico_w_asserts 1,025,380 -> 1,026,060 B used (+680), .bin 1,037,412 -> 1,038,092 (+680); pico_w_alpha 976,228 -> 976,228 B used (+0), .bin 990,380 -> 990,380 (+0); pico_w_air 1,019,480 -> 1,019,464 B used (-16), .bin 1,035,340 -> 1,035,340 (+0); pico2_w_release 999,072 -> 999,744 B used (+672), .bin 999,096 -> 999,768 (+672); pico_wire 592,260 -> 592,260 B used (+0), .bin 604,296 -> 604,296 (+0); pico_wiretft 774,772 -> 775,452 B used (+680), .bin 786,808 -> 787,488 (+680); pico_wiretft_angulo 814,700 -> 815,300 B used (+600), .bin 826,736 -> 827,336 (+600); pico2_w_angulo 1,035,992 -> 1,036,576 B used (+584), .bin 1,036,016 -> 1,036,600 (+584); pico_w_test_https 1,028,324 -> 1,029,020 B used (+696), no .bin recorded (ota_exempt) (main's #289-#295 merged into the branch: the English boot lists the pack, the Idioma choice waits for a boot, the guard arm; the changed functions are main's own pair, decoded against it). sync 6 C3: pico_wiretft_angulo 815,300 -> 815,356 B used (+56), .bin 827,336 -> 827,392 (+56); pico2_w_angulo 1,036,576 -> 1,036,632 B used (+56), .bin 1,036,600 -> 1,036,656 (+56) (the Angulo's Idioma screen lists the pack an English boot found (#294)). 2026-10-28 chore/test-image-license-fs: pico_w_test serves the license page from LittleFS too (custom_fs_pages LICENSE_PAGE), the maintainer's \"Abrir espaco na pico_w_test\": it signed 1,040,309 B against the OTA ceiling of 1,040,384 B, 75 B of room on main 9f4da77b, and the UI-drivers branch after its sync with main was 101 B over. -3,624 B (flash used 1,024,412, .bin 1,036,444, signed 1,036,685, 3,699 B of room; RAM unchanged at 109,212 B), not the 4,260 B of the page's gzip: no page of this image was served from the file before, so serveProtectedFsPage( ) (428 B; handleLicense( ) 8 B smaller) and the \"Page asset missing\" notice with the file name (212 B of strings in .rodata, 4 B of .text outside any symbol) are linked now; .text +424, .rodata -4,048. Its budget and \"bin\" come down. The other eight are byte for byte main's (9f4da77b), compared by sha256. sync 6b: pico_w_test 1,028,212 -> 1,024,580 B used (-3,632), .bin 1,040,244 -> 1,036,612 (-3,632) (main's #297 merged: pico_w_test serves its licence page from LittleFS (LICENSE_PAGE_GZ out, serveProtectedFsPage( ) in, handleLicense( ) 24 -> 16 B)). 2026-10-28 feat/web-theme-names on 99e9001b (main after #297): GET /api/themes sends each theme's code and the dashboard translates the stock names by it (formatThemeRow( ) in src/ThemeApiRow.h, themeOptions( ) in the TFT block of WebUI.h, the row buffer 96 B; each option carries data-en, the device's English, so that applyLang( ) restores it when /api/lang answered first). The six panel images grow: release +160 B (1,023,748, .bin 1,035,780; OTA room 4,523 -> 4,363), test +160 (1,024,572, .bin 1,036,604; 3,699 -> 3,539), test_https +168 (1,029,012; its .bin 1,041,044 is not recorded, ota_exempt), asserts +152 (1,026,044, .bin 1,038,076; 2,219 -> 2,067), pico2_w_release +168 (999,864, .bin 999,888) and pico_wiretft +160 (775,460, .bin 787,496). The three images with no panel do not link handleApiThemes( ), but its string literals sit in WebManager_Api.o's merged string pool, which the linker keeps whole: pico_wire +16 B (592,316, .bin 604,352), and pico_w_alpha (976,228, .bin 990,380) and pico_w_air (1,019,464, .bin 1,035,340) keep their sizes with another sha256, every symbol after the first changed string 16 B further on inside a .rodata of 434,176 and 454,656 B (multiples of 4,096), their page bytes identical (the page change of data-en is inside the TFT block and leaves them as they were). RAM unchanged on all nine. The same handler under #if SIMUT_DISPLAY_TFT, tried in a scratch tree, gave pico_wire 592,268 (-32 B against the budget it had) and was not taken: it would add an #if site to WebManager_Api.cpp. 2026-10-28 sync 7 (feat/ui-drivers + main c9028947 = #298, from scratch against the baseline 290c2662 built the same way): pico_w_release 1,023,764 -> 1,023,924 B used (+160), .bin 1,035,796 -> 1,035,956 (+160); pico_w_test 1,024,580 -> 1,024,740 B used (+160), .bin 1,036,612 -> 1,036,772 (+160); pico_w_asserts 1,026,060 -> 1,026,228 B used (+168), .bin 1,038,092 -> 1,038,260 (+168); pico_w_alpha 976,228 -> 976,228 B used (+0), .bin 990,380 -> 990,380 (+0); pico_w_air 1,019,464 -> 1,019,464 B used (+0), .bin 1,035,340 -> 1,035,340 (+0); pico2_w_release 999,744 -> 999,912 B used (+168), .bin 999,768 -> 999,936 (+168); pico_wire 592,260 -> 592,276 B used (+16), .bin 604,296 -> 604,312 (+16); pico_wiretft 775,452 -> 775,628 B used (+176), .bin 787,488 -> 787,664 (+176); pico_wiretft_angulo 815,356 -> 815,532 B used (+176), .bin 827,392 -> 827,568 (+176); pico2_w_angulo 1,036,632 -> 1,036,792 B used (+160), .bin 1,036,656 -> 1,036,816 (+160) (the theme-name page and route land on the branch: pico_w_release, pico_w_test, pico2_w_angulo +160, pico_w_asserts +168, pico2_w_release +168, pico_wiretft and pico_wiretft_angulo +176 (themeCleanCopy( ) +48 B, handleApiThemes( ) -20 B, the page's gzip +121..+129 B, the rows' strings, alignment), pico_wire +16 B (the strings), pico_w_alpha and pico_w_air the same size with other bytes; pico_w_test_https +168 B). 2026-10-28 sync 7, C2 (the Angulo's 18 theme names in English): pico_wiretft_angulo 815,532 -> 815,468 B used (-64), .bin 827,568 -> 827,504 (-64); pico2_w_angulo 1,036,792 -> 1,036,728 B used (-64), .bin 1,036,816 -> 1,036,752 (-64) (only the two Angulo images move, by the catalogue's strings in Temas.cpp (Portuguese names -> English); the nine products do not link it and are sha256-equal to the merge's). 2026-10-28 fix/web-upload-refused: doUpload( ) reads the answer of /api/upload, and a refusal gets a message of its own (fil_up_refused, English default in the page): FILE_PAGE_GZ 6,670 -> 6,779 B in pico_w_release, and +112 B measured and +112 B on the .bin in release, test, asserts, pico2_w_release, pico_wire and pico_wiretft. alpha and air carry the new page (6,779 B) and stay the same size, the alignment absorbing it; test_https carries no Files page and is unchanged. All nine rebuilt from scratch on c9028947 plus this change; main's own pico_w_release rebuilt first, at the recorded 1,023,748 B and a .bin of 1,035,780 B. 2026-10-30 feat/ui-drivers sync 8: pico_w_release 1,023,924 -> 1,024,028 B used (+104), .bin 1,035,956 -> 1,036,060 (+104); pico_w_test 1,024,740 -> 1,024,844 B used (+104), .bin 1,036,772 -> 1,036,876 (+104); pico_w_asserts 1,026,228 -> 1,026,332 B used (+104), .bin 1,038,260 -> 1,038,364 (+104); pico_w_alpha 976,228 -> 976,228 B used (+0), .bin 990,380 -> 990,380 (+0); pico_w_air 1,019,464 -> 1,019,464 B used (+0), .bin 1,035,340 -> 1,035,340 (+0); pico2_w_release 999,912 -> 1,000,024 B used (+112), .bin 999,936 -> 1,000,048 (+112); pico_wire 592,276 -> 592,388 B used (+112), .bin 604,312 -> 604,424 (+112); pico_wiretft 775,628 -> 775,740 B used (+112), .bin 787,664 -> 787,776 (+112); pico_wiretft_angulo 815,468 -> 815,580 B used (+112), .bin 827,504 -> 827,616 (+112); pico2_w_angulo 1,036,728 -> 1,036,840 B used (+112), .bin 1,036,752 -> 1,036,864 (+112) (the merge of main cf97b955 (#299, the Files page says a refused upload was refused; #300, SIMUT_VERSION 2.12.0; #301 and #302 touch no firmware input) into 22d83de3, all eleven rebuilt from scratch through the pio shim, the baseline 22d83de3 rebuilt the same way first and sha256-equal to sync 7's in all eleven. The only symbols that move are WebUI_GZ::FILE_PAGE_GZ (+109..+112 B) and LOGIN_PAGE_GZ (+1 B), as in main's own pair c9028947 -> cf97b955; .rodata moves 104 B in pico_w_release, pico_w_test and pico_w_asserts and 112 B in the other five page images (main's pair: 112 B in its six), the 8-byte alignment of the section absorbing the 110 B of symbols differently. pico_w_alpha and pico_w_air carry the new page and keep their size, pico_w_test_https carries no Files page and is unchanged). 2026-10-30 s10c/c4: pico2_w_release 1,036,840 -> 1,037,864 B used (+1,024), .bin 1,036,864 -> 1,037,888 (+1,024); pico_wiretft_angulo 815,580 -> 816,660 B used (+1,080), .bin 827,616 -> 828,696 (+1,080) (the Angulo's fonts and icons on /license (the @IF ui_angulo block, in the two Angulo images only) and on its licence screen (a heading and three paragraphs, four UA_TXT texts); the nine images that do not carry the Angulo keep their sha256). 2026-10-30 s10c/c4 review (sources pinned): pico2_w_release 1,037,864 -> 1,037,832 B used (-32), .bin 1,037,888 -> 1,037,856 (-32); pico_wiretft_angulo 816,660 -> 816,612 B used (-48), .bin 828,696 -> 828,648 (-48) (the Angulo's licence sources are a local table appended all or none (anexarFontesDoAngulo), smaller than the four guarded appends it replaces). 2026-10-30 s10c/c4 review (page labels): pico2_w_release 1,037,832 -> 1,037,848 B used (+16), .bin 1,037,856 -> 1,037,872 (+16); pico2_w_original 1,000,024 -> 1,000,040 B used (+16), .bin 1,000,048 -> 1,000,064 (+16); pico_wire 592,388 -> 592,388 B used (+0), .bin 604,424 -> 604,424 (+0); pico_wiretft 775,740 -> 775,740 B used (+0), .bin 787,776 -> 787,776 (+0); pico_wiretft_angulo 816,612 -> 816,612 B used (+0), .bin 828,648 -> 828,648 (+0); pico_w_release 1,024,028 -> 1,024,060 B used (+32), .bin 1,036,060 -> 1,036,092 (+32); pico_w_test 1,024,844 -> 1,024,844 B used (+0), .bin 1,036,876 -> 1,036,876 (+0); pico_w_asserts 1,026,332 -> 1,026,364 B used (+32), .bin 1,038,364 -> 1,038,396 (+32); pico_w_alpha 976,228 -> 976,228 B used (+0), .bin 990,380 -> 990,380 (+0); pico_w_air 1,019,464 -> 1,019,464 B used (+0), .bin 1,035,340 -> 1,035,340 (+0) (the FreeSans Bold row of the /license page names its four radio images instead of the five-image label, so the licence blob is 13 to 27 B longer on the radio images (pico_w_test and pico_w_test_https keep that page in LittleFS, pico_w_test_https has no budget row to move, and the Wire images keep their bytes)). 2026-10-30 fix/angulo-wifi-status: pico2_w_release 1,037,848 -> 1,037,856 B used (+8), .bin 1,037,872 -> 1,037,880 (+8) (uaVmStatus maps the Wi-Fi levels: the Angulo's status bar on the Pico 2 W release). 2026-10-30 fix/angulo-wifi-status r2: pico2_w_release 1,037,856 -> 1,038,032 B used (+176), .bin 1,037,880 -> 1,038,056 (+176); pico_wiretft_angulo 816,612 -> 816,764 B used (+152), .bin 828,648 -> 828,800 (+152) (the status bar's Wi-Fi levels: three Lucide icons (wifi-zero, wifi-low, wifi-high) in the shared registry, the shifted icon names, the new ladder in uaVmStatus( )). 2026-10-30 fix/angulo-wifi-status r3: pico2_w_release 1,038,032 -> 1,038,080 B used (+48), .bin 1,038,056 -> 1,038,104 (+48); pico_wiretft_angulo 816,764 -> 816,804 B used (+40), .bin 828,800 -> 828,840 (+40) (the status bar's Wi-Fi levels as Lucide's signal bars (four icons in, wifi-zero/low/high out of the registry), the cable without link in the neutral grey). 2026-10-30 fix/angulo-wifi-status r4: pico2_w_release 1,038,080 -> 1,038,112 B used (+32), .bin 1,038,104 -> 1,038,136 (+32); pico_wiretft_angulo 816,804 -> 816,860 B used (+56), .bin 828,840 -> 828,896 (+56) (the cable without link has an icon of its own, Lucide's unplug (one icon more in the registry, the ids after UNPLUG shift by one)). 2026-10-30 (S10c C6b): pico2_w_release 1,038,112 -> 1,038,080 B used (-32), .bin 1,038,136 -> 1,038,104 (-32); pico_wiretft_angulo 816,860 -> 816,836 B used (-24), .bin 828,896 -> 828,872 (-24) (the web-busy card and the accounts-full message say what the code does). 2026-10-30 S11 (the original panel's 14 units moved to src/ui/original/): pico_w_asserts 1,026,364 -> 1,026,356 B used (-8), .bin 1,038,396 -> 1,038,388 (-8); pico_wiretft 775,740 -> 775,732 B used (-8), .bin 787,776 -> 787,768 (-8); pico2_w_original 1,000,040 -> 1,000,032 B used (-8), .bin 1,000,064 -> 1,000,056 (-8) (link order only: .text loses 8 B of alignment padding on these three, the other eight images keep their sizes). 2026-10-31 fix/alarm-limits-snapshot on 3ca4976f (F3 of the v2.13.0 Known list): the limits editor's Salvar posts a copy into a 68 B member of DisplayManager, a TFT-only member, so the eight TFT images grow and the three LCD images do not. Used and .bin move together: pico_w_release +104, pico_w_test +104, pico_w_test_https +104, pico_w_asserts +104, pico2_w_release +48, pico_wiretft +144, pico_wiretft_angulo +104, pico2_w_original +120; pico_w_alpha, pico_w_air and pico_wire were built from zero at both commits and their firmware.bin has the same sha256. Each budget and .bin is set to the new measure, and the OTA table (PLANO_STABLE 3.1) says the same. 2026-10-31 fix/keypad-route-angulo on 6981fd93 (F4 of the v2.13.0 Known list): on the Angulo, GET /api/keypad sends the faces, the state, the policy and targets and no grid and no popup, chosen by a UI hook that is a constant on each side (uiKeypadTargets( ): nullptr on the original, \"/api/ui\" on the Angulo), so the branch an image does not take is not in the image: the two Angulo images shrink, pico2_w_release -208 and pico_wiretft_angulo -216 (used and .bin move together), and the other nine have the same sha256 as at 6981fd93. Both budgets and .bin are lowered to the new measure, and the OTA table (PLANO_STABLE 3.1) says the same. 2026-11-04 fix/small-bugs on 2128deb1 (main 34fc44a9 merged in, then twelve small fixes; all eleven built from zero through the pio shim of 2026-10-28 under the shared build lock, -Werror holding; main's records equal main's images on this host as of the same day): used against main pico_w_release +448, pico_w_test +480, pico_w_test_https +472 (used only), pico_w_asserts +440, pico_w_alpha +472 (.bin 0: the growth went into the linker fill in front of the Bluetooth bank), pico_w_air +408 (.bin 0: the growth went into the linker fill in front of the Bluetooth bank), pico2_w_release +336, pico_wire +104, pico_wiretft +64, pico_wiretft_angulo −24, pico2_w_original +432. The twelve fixes and the share of each are in CHANGELOG.md (Unreleased). Each budget and .bin is set to the new measure, and the OTA table (PLANO_STABLE 3.1) says the same. 2026-11-04 fix/small-bugs, recorded again at 212ff4bd after the review of the branch; used against the values this replaces: pico_w_release +0, pico_w_test +0, pico_w_test_https +0, pico_w_asserts +0, pico_w_alpha +0, pico_w_air +0, pico2_w_release +208, pico_wire +0, pico_wiretft +0, pico_wiretft_angulo +216, pico2_w_original +0. 2026-11-04 feat/alpha-lcd-live on 10c314c4 (main d8871388 and #319 merged in at d92b08a2, then steps 5a, 5b and the font's notice; all eleven built from zero through the pio shim of 2026-10-28 under the shared build lock, -Werror holding): the alpha's live LCD view (the glass that Core 1 publishes, GET /api/lcd, the box on the dashboard and its font) is in the two images with the alpha's LCD and in no other: the other nine have the same sha256 and size as main's images (tip2 table of 2026-11-04, pico_w_release 48c3b497 to pico2_w_original 4b066838), and the decoded pages inside each firmware.bin carry the box and the font's notice in pico_w_alpha and pico_wire only. Against the records it replaces: pico_w_alpha 976,732 -> 981,916 B used (+5,184 B) and .bin 990,380 -> 994,476 B (+4,096 B: one 4 KB step, the rest went into the linker fill in front of the Bluetooth bank), RAM 125,788 B; pico_wire 592,532 -> 597,372 B used (+4,840 B) and .bin 604,568 -> 609,408 B (+4,840 B), RAM 104,592 B. Steps 1 to 3 (the glass's shadow, GET /api/lcd) are +1,088 B used on the alpha and +1,296 B on the Wire, and the RAM of +108 B on both (measured at d92b08a2: alpha 977,820 B used, .bin 990,380 B; Wire 593,828 B, .bin 605,864 B); step 5 is the rest, +4,096 B on the alpha and +3,544 B on the Wire, of which the notice is +312 B on the Wire and 0 B on the alpha. Each budget and .bin is set to the new measure, and the OTA table (PLANO_STABLE 3.1) says the same. 2026-11-05 feat/alpha-lcd-live, recorded again at 03d05e47 after the page's glyph table became ROM_A00 of rom_hd44780.py (the maintainer's decision of 2026-11-04: the live view shows the characters the real LCD shows; the earlier font and its /license notice are gone) and after the review's fixes (the age in its own element, only a frame resets the count of failures, the body read with a timeout); all eleven built from zero through the pio shim of 2026-10-28 under the shared build lock, -Werror holding, and the nine without the alpha's LCD are sha256- and size-equal to main's images (tip2 table of 2026-11-04, 9 of 9). Used against the values this replaces: pico_w_alpha +0 (981,916 B; .bin 994,476 B, +0: its linker fill absorbs the change) and pico_wire -160 (597,372 -> 597,212 B; .bin 609,408 -> 609,248 B). On the Wire the notice's removal is -312 B and the new table's data +24 B (597,084 B at 9c456124), and the review's fixes +128 B. Against the records of main: pico_w_alpha +5,184 B used and +4,096 B .bin, pico_wire +4,680 B used and .bin, RAM 125,788 B and 104,592 B. The budget goes down with the image, and the OTA table (PLANO_STABLE 3.1) says the same. chore/v2.14.0, 2026-11-05: pico_w_release 1,024,828 -> 1,024,820 B used (-8), .bin 1,036,860 -> 1,036,852 (-8); pico_w_test 1,025,644 -> 1,025,636 B used (-8), .bin 1,037,676 -> 1,037,668 (-8); pico_w_asserts 1,027,124 -> 1,027,116 B used (-8), .bin 1,039,156 -> 1,039,148 (-8); pico_wire 597,212 -> 597,204 B used (-8), .bin 609,248 -> 609,240 (-8); pico_wiretft 776,164 -> 776,156 B used (-8), .bin 788,200 -> 788,192 (-8); pico_wiretft_angulo 817,140 -> 817,132 B used (-8), .bin 829,176 -> 829,168 (-8) (SIMUT_VERSION 2.13.0 -> 2.14.0, the only source change; the login page's gzip member packs 6,666 -> 6,660 B (LOGIN_PAGE_GZ, the only symbol whose size differs in the ELF of any of the eleven) and what follows it sits 8 B closer, so these six images and pico_w_test_https (no .bin record: 1,029,980 -> 1,029,972 B used, -8) lose 8 B, while pico_w_air, pico_w_alpha, pico2_w_release and pico2_w_original keep the same used and the same .bin, their padding absorbing the shift; all eleven built from zero through the pio shim under the shared build lock, -Werror holding, the unchanged tree built the same way first and sha256-equal to main's eleven (11 of 11)).; fix/web-texts-manual-review, 2026-11-06 (565f725c; all eleven from zero through the pio shim under the shared build lock, -Werror holding, no warning from src/; the base 95a1d2ae built the same way, the eleven .bin sizes read from the sibling session's wt_tags_base builds of that commit, equal the records to the byte; used against the base pico_w_release +512, pico_w_test +520, pico_w_test_https +424, pico_w_asserts +512, pico_w_alpha +192, pico_w_air +192, pico2_w_release +520, pico_wire +568, pico_wiretft +520, pico_wiretft_angulo +520, pico2_w_original +520, the .bin of pico_w_alpha and pico_w_air equal, their alignment absorbing the 192 B; the growth is the four fixes together (lang.js, four pages, 27 keys' markup, ScanResultJson.h), not split by fix; every budget and each changed image's .bin recorded); fix/web-texts-manual-review, the review's two commits, 2026-11-06 (32bb8a26: 38681b83 is test code only; 32bb8a26 gives the export box's failure counter a key of its own, exp_failed, in place of the dashboard's dash_metr_err; all eleven from zero again through the pio shim under the shared build lock, -Werror holding, no warning from src/; used against the first build of this branch (565f725c) pico_w_release -8, pico_w_test -8, pico_w_test_https -8, pico_w_asserts -8, pico_wire -8, pico_wiretft -8, pico2_w_original -8, and pico_w_alpha, pico_w_air, pico2_w_release and pico_wiretft_angulo equal; against the base 95a1d2ae the growth is pico_w_release +504, pico_w_test +512, pico_w_test_https +416, pico_w_asserts +504, pico_w_alpha +192, pico_w_air +192, pico2_w_release +520, pico_wire +560, pico_wiretft +512, pico_wiretft_angulo +520, pico2_w_original +512; the seven budgets and .bin that moved recorded, the budgets lowered with the images); fix/web-texts-manual-review, the history page's two fixes, 2026-11-06 (246a2a7c: the Origem selector, a count element of its own and a key on the label only while it shows a word; eacef7be: the Max and Min badges, hist_max and hist_min in both packs, which are files on the device and not part of the image; all eleven from zero again in one pio run through the pio shim under the shared build lock, -Werror holding, no warning from src/; used against the review's build (50a78f05) pico_w_release +160, pico_w_test +160, pico_w_test_https +160, pico_w_asserts +160, pico_wire +160, pico_wiretft +160, pico_wiretft_angulo +152, pico2_w_original +136, pico2_w_release +128, and pico_w_alpha and pico_w_air equal: their .rodata is a multiple of 4,096 B (438,272 and 454,656), so the padding absorbs the page's growth, and the new page is in both ELFs (the history page's gzip member read back out of each, with the new markup in it); against the base 95a1d2ae the growth is pico_w_release +664, pico_w_test +672, pico_w_test_https +576, pico_w_asserts +664, pico_w_alpha +192, pico_w_air +192, pico2_w_release +648, pico_wire +720, pico_wiretft +672, pico_wiretft_angulo +672, pico2_w_original +648; the nine budgets and .bin that moved recorded); fix/web-texts-manual-review, the second review's commits, 2026-11-06 (778cd368: the Origem label's count guard; c674df49: the overlay and the progress lines of the history page written through _setText( ), with their key and English or none; all eleven from zero again in one pio run through the pio shim under the shared build lock, -Werror holding, no warning from src/; used against the history page's first two fixes (eacef7be, recorded in 1b2b4785) pico_w_release +72, pico_w_test +72, pico_w_test_https +72, pico_w_asserts +72, pico2_w_release +104, pico_wire +72, pico_wiretft +72, pico_wiretft_angulo +80, pico2_w_original +96, and pico_w_alpha and pico_w_air equal (their .rodata is still a multiple of 4,096 B, 438,272 and 454,656, which absorbs the growth; the new page is in both ELFs, read back out of each, with _setText( ) in it); against the base 95a1d2ae the growth is pico_w_release +736, pico_w_test +744, pico_w_test_https +648, pico_w_asserts +736, pico_w_alpha +192, pico_w_air +192, pico2_w_release +752, pico_wire +792, pico_wiretft +744, pico_wiretft_angulo +752, pico2_w_original +744; the nine budgets and .bin that moved recorded); feat/screen-boot-tags, 2026-11-06, on 95a1d2ae (before the merge of main): pico_w_test 1,025,636 -> 1,027,884 B used (+2,248), .bin 1,037,668 -> 1,039,916 (+2,248), RAM 109,212 -> 111,060 B (+1,848); pico_w_test_https 1,029,972 -> 1,032,220 B used (+2,248), RAM 109,228 -> 111,076 B (+1,848) (the boot terminal's recorder, `screen boot` and `screen clkboot`, all under SIMUT_BOOT_TAGS, SIMUT_CLI_FULL && SIMUT_DISPLAY_TFT: .text +2,056, .rodata +192, .bss +1,848, .data unchanged; the other nine images sha256-equal to 95a1d2ae, all eleven built from zero through the pio shim on both sides, base and tip each in a worktree of its own, and each one's .bin equal to its record before this change).; feat/screen-boot-tags after the merge of main (#324), 2026-11-06: pico_w_test built from zero through the pio shim under the shared build lock, -Werror holding, no warning from src/: 1,028,628 B used (main 1,026,380, +2,248), .bin 1,040,660 B (main 1,038,412), RAM 111,060 B; the budget recorded as measured and the 'bin' dropped with the exemption; the nine images without the tags, built the same way, equal main's records to the byte in used and in .bin.; feat/screen-boot-tags with the tags in pico_w_test only (019a4996), 2026-11-06: all eleven from zero in one pio run through the pio shim under the shared build lock, -Werror holding, no warning from src/, and a clean copy of main (3b09e9ff) built the same way beside them; the ten images without the tags are sha256-equal to main's, pico_w_test_https included (1,030,620 B used, 1,042,652 B .bin, as main; with the tags it did not link, 420 B over the program slot), and pico_w_test is +2,248 B used and +2,248 B .bin against main (1,028,628 and 1,040,660 B); no budget and no 'bin' moves.; chore/repo-hygiene-2026-11-07 on 6c8fff60 plus this change, 2026-11-07 (all eleven built from zero through the pio shim of 2026-10-28 against ~/.platformio, no other build running, -Werror holding, no warning from src/: the login page's tagline gains 'Integrado' and the Angulo's boot subtitle changes; used against 6c8fff60 +8 on pico_w_release, pico_w_test, pico_w_test_https, pico_w_asserts, pico_wire, pico_wiretft and pico2_w_original, the .bin of each +8; pico_w_alpha, pico_w_air, pico2_w_release and pico_wiretft_angulo equal in used and in .bin; the seven budgets and five .bin recorded)",
+ "measured_on": "2026-10-05 chore/v2.11.1: SIMUT_VERSION 2.11.1, the only source change. All seven rebuilt here. The version has as many characters as 2.11.0, but the login page carries it, and its gzip came out 5 B smaller in every image: measured and .bin 8 B under the record on release, test, asserts and pico2_w_release; alpha, air and test_https equal, their alignment absorbing it. Budgets unchanged. 2026-10-12 feat/lan-switch: pico_wire joins (the SIMUT Wire, a plain Pico with the network on an ENC28J60, no driver yet), measured 587,980 B and a .bin of 600,016 B; the seven others are byte for byte what main 2a391da builds. Then twelve sensor slots (-DMAX_SENSORS=12, GP12-GP15 are the module's): 588,020 B and a .bin of 600,056 B, +40 B for clearing the slots past the twelfth on every config load; RAM 102,416 -> 102,064 B. Then eleven (the module's INT moved to GP11, 2026-10-12): 587,972 B and a .bin of 600,008 B, RAM 101,980 B, measured on 28a54c8. 2026-10-13 feat/wire-driver: the wired driver takes pico_wire to 597,448 B and a .bin of 609,480 B (RAM 104,680 B); the seven Wi-Fi images stay byte for byte what dd7efdc builds until the top-panel fix, which grows them by 32 to 64 B (release, test, test_https and asserts +48, alpha +64, air +32, pico2_w_release +56; .bin +48 on release, test and asserts, +56 on pico2_w_release, unchanged on alpha and air). All eight measured at 4912762's tree plus this change. The review's fixes (module before the sensors, the static address checks, the MAC before the start, syslog and the OTA teardown on the cable, the log latches): pico_wire 598,560 B (+1,112 B), .bin 610,592 B, RAM 104,688 B; the seven Wi-Fi images compile none of it. 2026-10-13 feat/wire-surface: pico_wiretft joins (the SIMUT Wire with the TFT panel and touch): 780,628 B and a .bin of 792,664 B (RAM 109,784 B), built on the driver's tree plus the manifest entry and no change to src/. The eight others keep their records: no flag and no source of theirs moved. Then the surface a wired build loses with the radio (the web pages, the console, the panels and the third-party list): pico_wire 595,632 B and a .bin of 607,664 B (RAM 104,684 B), pico_wiretft 777,428 B and a .bin of 789,464 B (RAM 109,788 B), both budgets brought down to them; the seven radio images byte for byte the same at each of the branch's commits, compared by sha256 (pico_w_release 8b565f66 throughout). Then the captive-portal redirect and probes, the setup AP's, compiled out of the wired builds: pico_wire 595,016 B (.bin 607,048), pico_wiretft 776,820 B (.bin 788,856), RAM unchanged; budgets brought down again. Then four log codes of the cable's own (529-532) with their three rows in the edge filter, compiled into the wired builds only: pico_wire 595,152 B (+136, .bin 607,184), pico_wiretft 776,956 B (+136, .bin 788,992); the seven radio images unchanged. 2026-10-13 feat/net-boundary: the wired link generic over LwipIntfDev (the chip in a device header, src/net/Enc28j60Device.h) and a module stopped by an update that did not happen started again by the retry: pico_wire 595,208 B (+56, .bin 607,240), pico_wiretft 777,004 B (+48, .bin 789,040), RAM unchanged. The move alone is byte for byte neutral (pico_wire sha256 3edd1440, pico_wiretft e58d65e9, before WiredLink::chipName( ) existed); the growth is the module-fail message built from the driver's name (a second concatenation in startWired( ), plus chipName( ) and started( ) now reached) and the restart path in wiredLost( ), the only two function bodies that differ in the normalized disassembly of both images; the seven radio images are byte for byte the same. Then the log line of that path names what happened (\"module stopped\", where it said \"link down\"): +48 B on both, pico_wire 595,256 B (.bin 607,288), pico_wiretft 777,052 B (.bin 789,088); the seven radio images unchanged. Then the review of that change: the stopped module is looked for first in updateWired( ), in every state, because a unit still on its way up (NET_CONNECTING: booted with the cable out, or a lease pending) never reached wiredLost( ) and was never started again: +40 B on both, pico_wire 595,296 B (.bin 607,328), pico_wiretft 777,092 B (.bin 789,128), RAM unchanged; updateWired( ) and wiredLost( ) the only two function bodies that differ in the normalized disassembly of both images. The seven radio images are byte for byte the same, pico_w_air included: its radio power-down moved into net::powerDownForSleep( ) (src/net/NetLink.h) without moving a byte. 2026-10-13 feat/net-radio-unit: the radio's state machine moved from NetworkManager.cpp into its own unit, src/net/NetworkManager_Cyw43.cpp, which the wired images do not build, and NetworkManager.cpp became the host of what every network needs, reaching the link through three methods each unit defines. The wired images lose the radio code they compiled dead: pico_wire 590,740 B (-4,556, .bin 602,776), pico_wiretft 772,852 B (-4,240, .bin 784,888), both budgets brought down. The radio images pay for the split: release +48 B (1,021,180, .bin 1,033,212), asserts +40 (1,023,468, .bin 1,035,500), pico2_w_release +120 (997,328, .bin 997,352), test_https -8 (1,030,692), test, alpha and air the same; the normalized disassembly of all seven differs in four bodies only: begin( ) and update( ) (now the host's), handleConnecting( ) and getRssi( ), which call resetReconnectLadder( ) and isConnected( ) across the unit boundary where the compiler inlined them before. Inlining the two helpers in the header instead was measured and grew the radio images more (release +88, test +56, asserts +80). 2026-10-13 fix/radio-link-in-ntp-wait: the radio watches its link inside the NTP wait, one check in its updateLinkFirst( ) and NET_READY's loss in a helper that logs it, wifiLost( ). The seven radio images grow: release +72 B (1,021,252, .bin 1,033,284), test +80 (1,025,788, .bin 1,037,820, under its budget), test_https +80 (1,030,772), asserts +80 (1,023,548, .bin 1,035,580), alpha +88 (975,060, under its budget, the .bin the same), air +80 (1,018,336, under its budget, the .bin the same), pico2_w_release +72 (997,400, .bin 997,424); the two wired images are byte for byte the same. With the log at each of the two callers instead of in the helper, release grew +96 B. 2026-10-13 fix/silent-radio-no-blind-join (#266): a reconnect scan whose start the radio did not answer, and that never finished, is not escaped by a blind join; two in a row past the first half hour restart the radio by safeReboot( ). The radio images grow: release +376 B (1,021,628, .bin 1,033,660), test +352 (1,026,140, .bin 1,038,172), test_https +352 (1,031,124), asserts +384 (1,023,932, .bin 1,035,964), alpha +280 (975,340) and air +280 (1,018,616), both .bin the same, pico2_w_release +376 (997,776, .bin 997,800). The wired images carry the two new members of the shared header and the host's reset of one: pico_wire the same size (590,740), pico_wiretft +16 (772,868, .bin 784,904). Writing the two log lines with one snprintf each instead of the String concatenations was measured: 16 B less on release, not worth the buffer. 2026-10-14 fix/link-health-in-ntp-wait: getRssi( ) answers the level of any link that is up, the NTP wait included, on the radio and on the cable (isLinkUp( ) where it read isConnected( )). The RP2040 images are the same size; pico2_w_release +8 B (997,784, .bin 997,808). With it, the Air's FLUSH waits for NTP 10 s at most (AIR_FLUSH_NTP_WAIT_MS, airFlushNetLost( )): pico_w_air +48 B (1,018,664, the .bin the same), the other eight byte for byte the same. 2026-10-14 fix/web-short-write: override 2m (webserver_short_write.patch), a write cut short ends the response and closes the client as its real type (stopClient( ), so over TLS too), and the count read out (metr.cgw, /metrics, and `show net status` with the TCP segment pool and tcp_write's memory failures). Every image grows: release +512 B (1,022,140, .bin 1,034,172), test +480 (1,026,620, .bin 1,038,652), test_https +504 (1,031,628), asserts +504 (1,024,436, .bin 1,036,468), alpha +344 (975,684) and air +352 (1,019,016), both .bin the same, pico2_w_release +520 (998,304, .bin 998,328), pico_wire +480 (591,220, .bin 603,256), pico_wiretft +504 (773,372, .bin 785,408). Of the release's 512 B, 394 are sized symbols (the status handler 72, _writeAll( ) and _shortWrite( ) 78, the console 56, the metrics 32, the three _currentClientStop( ) 32, the framework's call sites and WebManager the rest) and the rest the new strings. 2026-10-14 fix/web-fewer-writes: override 2n (webserver_fewer_writes.patch), HTTPServer collects a response's small writes in a buffer of one TCP segment and sends them together (#189). Every image grows: release +328 B (1,022,468, .bin 1,034,500), test +328 (1,026,948, .bin 1,038,980, 1,163 B of OTA headroom left), test_https +328 (1,031,956), asserts +328 (1,024,764, .bin 1,036,796), alpha +328 (976,012, the .bin the same), air +328 (1,019,344) with its .bin +4,096 (1,035,340), pico2_w_release +296 (998,600, .bin 998,624), pico_wire +328 (591,548, .bin 603,584), pico_wiretft +328 (773,700, .bin 785,736). The Air's .bin moves in 4 KB steps, as the alpha's does: their .rodata starts on a 4 KB boundary because it holds BTstack's __bluetooth_tlv (8 KB, sector-aligned), and the Air's .text ended 196 B short of the next step. Against main's release the sized symbols grow 362 B and the image 328: _outAppend( ) 140, the constructor's malloc( ) 56, _outBodySent( ) 32, _outFlush( ) 22, and the flushes and resets in send( ), sendContent( ), _prepareHeader( ), _handleRequest( ), _finalizeResponse( ) and the destructor the rest. The 1,460 B buffer is heap, allocated with the server. As an array among the class's members instead of behind a pointer at its end, it pushed every member after it past the Cortex-M0+'s short load/store offsets: release +1,152 B, against +352 B for the same code behind the pointer. The bench ran the release of the branch's first version (sha256 b1a7e7da), which had two more stores, _outLen = 0 in _shortWrite( ), which nothing reaches, and in stopClient( ), which runs but changes nothing: what it would drop goes to a client already closed, whose writes return 0. The final normalized disassembly differs from the bench image's in _shortWrite( ) and dropAbortedStream( ) only, 16 B. 2026-10-15 deps/pico-enc28j60-1.3.1: the ENC28J60 library moves from fb16ec1 to the commit of v1.3.1 (4984e2f), whose lwIP interface keeps up to eight frames in the chip and uploads one while the one before is on the wire (the library's 1.3.0; 1.3.1 changed an example and the documentation only). The wired images grow: pico_wire +568 B (592,116, .bin 604,152), pico_wiretft +624 (774,324, .bin 786,360), RAM +56 on both (the queue's state in the driver object). Nearly all of it is the new send path (_sendQueued( ) 228, _sendPoll( ), _sendFinish( ) and enc28j60_frame_send_finish_blk( ) 100 each, _sendStartNext( ) 80, against sendFrame( ) -144 and enc28j60_frame_send_blk( ) -90); the /license page, whose third-party note names the new commit, 12. The seven radio images are byte for byte main's (192e1ad), compared by sha256. 2026-10-15 fix/wire-tft-flash-render-274: the WiFiServerSecure accept override (wifiserversecure_drop_plaintext.patch, patch.sh 2b-ter) drops a non-TLS client on the TLS port before the BearSSL handshake, decided from its first byte (#274). It is compiled wherever SIMUT_WEB_HTTPS is set, so the five HTTPS images grow: pico_wiretft +56 B (774,380, .bin 786,416), release +48 (1,022,516, .bin 1,034,548), test_https +48 (1,032,004), asserts +56 (1,024,820, .bin 1,036,852), pico2_w_release +24 (998,624, .bin 998,648). pico_wire, pico_w_test, alpha and air compile no HTTPS server, so the patched accept path is not in them and they keep main's sizes (d5f7063). 2026-10-15 perf/tft-spi-clk-peri: setup( ) clocks clk_peri from clk_sys on the six images that drive the panel and SIMUT_TFT_SPI_HZ is 66.5 MHz; /api/status gains sys.clk_peri and sys.tft_hz through macros, so the images without the panel compile the very tokens they did. Flash used (.bin): release +216 (1,022,732, .bin 1,034,764), test +208 (1,027,156, .bin 1,039,188), test_https +208 (1,032,212), asserts +208 (1,025,028, .bin 1,037,060), pico2_w_release +224 (998,848, .bin 998,872), wiretft +208 (774,588, .bin 786,624). pico_wire, alpha and air are byte for byte main's (80c3c88), compared by sha256. 2026-10-14 feat/ui-drivers (S2b-3): pico_wiretft_angulo joins (the Ângulo UI driver's bench image: pico_wiretft with `ui = \"angulo\"`, the touch panel drawn by src/ui/angulo/, transitional and not approved for products): 819,528 B and a .bin of 831,560 B (RAM 116,292 B), +45,828 B and +45,824 B over pico_wiretft (773,700 B, .bin 785,736, RAM 109,784 B), budget = measured. Built against a PlatformIO core that carries overrides 2m and 2n, the private copy that reproduces CI to the byte: in the same session pico_wiretft came out at 773,700 B (.bin 785,736) and pico_w_release at 1,022,468 B (.bin 1,034,500), equal to their records. The image is not exempt from the OTA ceiling: it takes the bench OTA (ota_trust_bench), and signed it is 831,801 B, 208,583 B under 1,040,384; its row is in the OTA headroom table of PLANO_STABLE.md section 3.1. The nine others keep their records. 2026-10-15 feat/ui-drivers-s5 (S5): the Painel group's touches become core actions (src/DisplayManager_Actions.h). pico_w_release, pico_w_test and pico_w_test_https grow 8 B at C3 and pico_w_asserts 8 B at C6 (each of the two steps adds two instructions to handleTouch in all four: 1,022,404, 1,026,884, 1,031,892 and 1,024,708 B used; .bin 1,034,436, 1,038,916 and 1,036,740 B), measured on 60eabd6 (the C8 builds) and again on 33836c2 (the C9 builds), the same sizes and the same sha256 for the nine product images, built against the private core that carries overrides 2m and 2n; pico2_w_release, pico_wiretft, pico_w_air, pico_w_alpha and pico_wire are the size they had at the end of S4. pico_wiretft_angulo 815,424 -> 816,264 B used (+840 B: C2 -32, C3 -88, C4 +136, C5 +328, C6 +424, C7 +72, C8 0), .bin 828,296 B. The review of S5 (C10c): `uiBootSkipHit` asks which frame is on the glass before it reads the frame's targets (`s_modo == MODE_DASHBOARD`, one comparison more), so another screen's target of type 1 cannot read as the boot's skip button. Only pico_wiretft_angulo compiles that file: 816,264 -> 816,280 B used (+16: `uiBootSkipHit` 29 -> 34 instructions, and one padding instruction in `_fini`, the only two functions that differ in the normalized disassembly of the image), .bin 828,296 -> 828,312 B, budget = measured; its row in the OTA headroom table of PLANO_STABLE.md moves with it (211,847 -> 211,831 B of headroom). Measured on a1fda61 plus this change, built against the same private core. 2026-10-15 feat/ui-angulo-visual (the panel's look: its own colour tokens, smoothed corners, the 9 pt text floor), merged over S5 at 2c89d12: pico_wiretft_angulo 816,280 -> 816,104 B used (.bin 828,312 -> 828,136), -176 B: the 512 B the smoothed corners took out on their own (8411d14, measured on the end of S4, which recorded only the numbers; Adafruit GFX's arc helpers are no longer linked) less what the 9 pt layouts added. Then the delivery's SANS7 and SANSB7 leave the kit, which nothing draws since the floor (4,939 B of glyph tables and bitmaps): 811,064 B used, .bin 823,096 B (-5,040 B), budget = measured; its row in the OTA headroom table of PLANO_STABLE.md moves to 217,047 B. The nine product images do not compile src/ui/angulo/ and keep their records. Built against the private core that carries overrides 2m and 2n. 2026-10-15 feat/ui-drivers-s6 (S6, the Base and Menus groups' touches as core actions): the nine product images are byte-equal to the end of S5 at every step and keep their records. pico_wiretft_angulo 811,064 -> 811,232 B used (+168 B: C1 7ee79e9 0, C2 a872d93 +128, C3 77e89aa -24, C4 390fc6f 0, C5 05da4f0 0, C6 80ecdb0 +64, C7 5da8888 0, C8 c078ef4 0), .bin 823,096 -> 823,264 B, budget = measured; its row in the OTA headroom table of PLANO_STABLE.md moves to 216,879 B. C3's -24 B carries a deviation: once actUserDelConfirm( ) added a call site in uaToqueBase( ), GCC stopped inlining pushUiEvent( ) into the Angulo's uiPass( ) (714 -> 674 instructions, one more bl; the out-of-line pushUiEvent( ) is unchanged and only this transitional image is affected; stages/s6_c3/accept_main_session.txt, UI_DRIVERS.md section 2.9). C7 and C8 change no size: C7 moves utf8FoldNext( ) and its three switch tables to an earlier place in the image (no instruction changes; stages/s6_c7/constructor_proof.txt), and C8's src/display/ListPager.h is included by the Angulo's glue only. Built against the private core that carries overrides 2m and 2n. 2026-10-15 feat/ui-drivers-s7 (S7, the Gráficos group's touches as core actions): the nine product images keep their sizes, budgets and OTA rows; C4 64dc6c6 changed the bytes of five of them, in handleTouch( ) only and at the same instruction count (pico_w_release, pico_w_test, pico_w_test_https, pico_w_asserts and pico_wiretft; their hashes are in stages/s7_c4/sha256.txt), and the other four are byte for byte what S6 left. pico_wiretft_angulo 811,232 -> 811,304 B used (+72 B: C1 7fb7d18 0, C2 73ecbd9 0, C3 43960a7 +80, C4 64dc6c6 -8, C5 af96f64 0, C6 6102e9f 0, C7 5e27b04 0), .bin 823,264 -> 823,336 B, budget = measured; its row in the OTA headroom table of PLANO_STABLE.md moves to 216,807 B. C3: handleTouch( ) 4,581 -> 4,603 instructions and uaToqueGraficos( ) 322 -> 346, with drawGraphHeaderBar( ) out of the image; C4: 4,603 -> 4,609 and 346 -> 343, with drawGraphBusyHint( ) out. C6 and C7 change no size: C6 permutes 28 linker veneers among themselves (231 bytes of the .bin, every one a call to a veneer or the veneer's slot; stages/s7_c6/layout_proof.txt) and C7's uaToqueGraficos( ) is two instructions shorter (343 -> 341) inside an image that keeps its size. Built against the private core that carries overrides 2m and 2n.2026-10-15 fix/ui-angulo-aviso: the status bar's badge finds its room between the brand and the icons (src/ui/angulo/telas/Telas.cpp), and the view model carries the notice's kind, seconds and name instead of a composed string; only pico_wiretft_angulo changes, 811,232 -> 811,656 B (+424 B: the width arithmetic, the cut of the name and the words moved from the glue into desenharStatus( )), .bin 823,264 -> 823,688. The nine product images do not compile src/ui/angulo/ (git diff --stat touches only src/ui/angulo/, tools/ and docs/). Built against the private core that carries overrides 2m and 2n; not run on a board. Then the web notice reads \"Web: \", the original top bar's form (its fixed words are 42 px where \" entrou pela web\" took 136, so a 15-character name shows whole beside the cable icon): 811,656 -> 811,640 B (-16 B, the budget comes down with it), .bin 823,688 -> 823,672; the nine product images are still untouched. 2026-10-15, the notice fix merged over S7 (fix/ui-angulo-aviso with be61a2f): pico_wiretft_angulo 811,712 B used, .bin 823,744 B, budget = measured; the nine product images do not compile src/ui/angulo/ and keep S7's records. 2026-10-15, fix/ui-angulo-toque on a6026d5 plus this change (touch targets that answer a finger): pico_wiretft_angulo 811,712 -> 811,928 B (+216 B: the hit test's second pass, the nearest box within 12 px, and the wider rectangles of the title bar's actions and of the limits pager's arrows; not split between the two, built once), .bin 823,744 -> 823,960 B, budget = measured; the nine product images do not compile src/ui/angulo/ (git diff --stat touches only src/ui/angulo/, tools/ and docs/) and keep their records. Built against the private core that carries overrides 2m and 2n, no patch.sh run; not run on a board. 2026-10-15, fix/ui-angulo-historico (a6026d5 plus this change): pico_wiretft_angulo 811,736 B used (+24 B), .bin 823,768 B (+24), budget = measured: graphAnchorStep( ) and graphOpenClearsAnchor( ) in Core 0 under SIMUT_UI_ANGULO; the nine product images do not compile that code (the `#if` is false in them) and keep S7's records. 2026-10-15, the two merged on fix/ui-angulo-navegacao: pico_wiretft_angulo 811,952 B used, .bin 823,984 B, budget = measured (the sum of both, +216 and +24). 2026-10-15, fix/ui-angulo-navegacao (the inert targets, the central band, the page cycle, the month arrows; 2aa380a plus this change, built once): pico_wiretft_angulo 811,952 -> 812,352 B used (+400 B), .bin 823,984 -> 824,384, budget = measured; the targets the screens list for their disabled controls, the band and its page cycle in the glue, and alvoDoToque( ); the control recorder (UA_CONTROLE( )) is ( (void)0 ) on the device. The nine product images do not compile src/ui/angulo/ (git diff --stat touches only src/ui/angulo/, tools/ and docs/) and keep their records. Built against the private core that carries overrides 2m and 2n, no patch.sh run; not run on a board. 2026-10-15, fix/ui-angulo-navegacao, the review's fixes (the host check's pixel rule and the recorder's test, the title as an inert target, the detail's page rule; 67c6521 plus this change, built once): pico_wiretft_angulo 812,352 -> 812,544 B used (+192 B: one more inert target on every screen with a title bar, and the glue's page rule), .bin 824,384 -> 824,576, budget = measured; the host-only code (the control recorder, the drop counter) is ( (void)0 ) or absent on the device. The nine product images do not compile src/ui/angulo/ (git diff --stat touches only src/ui/angulo/, tools/ and docs/) and keep their records. Built against the private core that carries overrides 2m and 2n, no patch.sh run; not run on a board. 2026-10-15, fix/ui-angulo-historico, second version (4122282 plus this change: the live base replaces the half-step bound): pico_wiretft_angulo 811,792 B used (+56 B on 4122282, +80 B on a6026d5), .bin 823,824 B (+56), budget = measured: graphAnchorStep( ) with the base, graphAnchorDay( ) and graphAnchorOpen( ) in Core 0 under SIMUT_UI_ANGULO; the nine product images do not compile that code and keep S7's records. 2026-10-15, fix/ui-angulo-historico (d2bf3cd, 456a1ad) merged over fix/ui-angulo-navegacao 91df450: pico_wiretft_angulo 812,600 B used, .bin 824,632 B, budget = measured. 2026-10-15, feat/ui-drivers-s8a (S8a, C1 to C8 over 9bef346, no patch.sh run, not run on a board): the nine product images moved in handleTouch( ) at C2 to C5 and kept their sizes but for C5, which took 8 B off six of them (pico_w_release 1,022,396 B used, .bin 1,034,428; pico_w_test 1,026,876, .bin 1,038,908; pico_w_test_https 1,031,884; pico_w_asserts 1,024,700, .bin 1,036,732; pico2_w_release 998,528, .bin 998,552; pico_wiretft 773,628, .bin 785,664); pico_wiretft_angulo 812,600 -> 813,048 B used (+448: C2 -48, C3 +144, C4 +320, C5 -8, C6 +40, C7 and C8 0), .bin 824,632 -> 825,080, budget = measured. Each action is compiled twice in that image (inlined into handleTouch( ) and into its own glue) until S10b. 2026-10-15, fix/ui-angulo-margem (3ba3ea5 plus this change, no patch.sh run, not run on a board): the panel's 4 px margin is black in both themes (ua::margemPreta( ) after each strip's draw in uaPintar( ), and the glass cleared to black on a full repaint): pico_wiretft_angulo 813,048 -> 813,152 B used (+104 B), .bin 825,080 -> 825,184, budget = measured; its row in the OTA headroom table of PLANO_STABLE.md moves to 214,959 B. The nine product images do not compile src/ui/angulo/ and keep their records (git diff --stat touches only src/ui/angulo/, tools/ and docs/). 2026-10-15 perf/ui-angulo-pintura (p1, a frame is drawn once and replayed): pico_wiretft_angulo 813,152 -> 814,496 B (+1,344), .bin 825,184 -> 826528; RAM 116,300 -> 131,156 B (+14,856, the record: 1024 calls of 14 B and 512 B of text); the nine product images do not compile src/ui/angulo/. 2026-10-15 perf/ui-angulo-pintura (p2, the strip hash by words): pico_wiretft_angulo 814,496 -> 814,664 B (+168), .bin 826,528 -> 826,696; RAM 131,156 -> 131,156 B (+0). 2026-10-15 perf/ui-angulo-pintura (p3, the corner and disc coverage tables): pico_wiretft_angulo 814,664 -> 816,792 B (+2,128), .bin 826,696 -> 828,824; RAM 131,156 -> 131,156 B (+0). 2026-10-15 perf/ui-angulo-pintura (p4, glyphs straight into the strip): pico_wiretft_angulo 816,792 -> 817,208 B (+416), .bin 828,824 -> 829,240; RAM 131,156 -> 131,156 B (+0). 2026-10-15 perf/ui-angulo-pintura (p5, fills by words): pico_wiretft_angulo 817,208 -> 817,256 B (+48), .bin 829,240 -> 829,288; RAM 131,156 -> 131,156 B (+0). 2026-10-15 perf/ui-angulo-pintura (p6, no black frame): pico_wiretft_angulo 817,256 -> 817,016 B (-240), .bin 829,288 -> 829,048; RAM 131,156 -> 131,156 B (+0). 2026-10-15 perf/ui-angulo-pintura (p7, the record in the canvas's spare rows, the chart's plot as three repeatable calls): pico_wiretft_angulo 817,016 -> 817,448 B (+432), .bin 829,048 -> 829,480; RAM 131,156 -> 116,348 B (-14,808). 2026-10-15 perf/ui-angulo-pintura (p8, the strips pushed by DMA while the next is composed, in src/ui/angulo/DisplayManager_UiA_Empurrar.cpp; the part of a strip on the glass at an alignment offset too): pico_wiretft_angulo 817,448 -> 817,424 B (-24), .bin 829,480 -> 829,456; RAM 116,348 -> 116,376 B (+28). 2026-10-16 perf/ui-angulo-pintura (p9, the review's fixes: the record's list returned after a paint, the in-flight flag set before the bus opens): pico_wiretft_angulo 817,424 -> 817,432 B (+8), .bin 829,456 -> 829,464; RAM 116,376 -> 116,376 B (+0). 2026-10-16 perf/ui-angulo-pintura (p10, line( ), fillRect( ) and rect( ) drop what lies outside the strip's window): pico_wiretft_angulo 817,432 -> 817,504 B (+72), .bin 829,464 -> 829,536; RAM 116,376 -> 116,376 B (+0). 2026-10-16 perf/ui-angulo-pintura (p11a, the plot spares the buckets a strip cannot show): pico_wiretft_angulo 817,504 -> 817,920 B (+416), .bin 829,536 -> 829,952; RAM 116,376 -> 116,376 B (+0). 2026-10-16 perf/ui-angulo-pintura (p11b, rounded boxes: runs as spans, in-place blend, body in three pieces): pico_wiretft_angulo 817,920 -> 818,200 B (+280), .bin 829,952 -> 830,232; RAM 116,376 -> 116,376 B (+0). 2026-10-16 perf/ui-angulo-pintura (p12a, the strip's fill and hash run from RAM): pico_wiretft_angulo 818,200 -> 818,216 B (+16), .bin 830,232 -> 830,248; RAM 116,376 -> 116,744 B (+368). 2026-10-16 sync/ui-drivers-main (main 2d6650b merged into feat/ui-drivers 1a913a4, so that the branch carries #275 PicoENC28J60 1.3.1, #276, #277 and #278): every image rebuilt on the merged tree against the default PlatformIO core (~/.platformio, framework 1.50601.0), which carries every override of patch.sh, 2b-ter (the WiFiServerSecure accept patch of #277) included; pico_w_release was built a second time at the end and came out equal. The first measurement of this tree, on the private core of the earlier entries, which lacks 2b-ter, is replaced: pico_w_test, pico_w_alpha, pico_w_air and pico_wire are byte for byte the same on both cores, and the HTTPS images grew by what main measured for that patch (pico_wiretft +56 B, release +48, test_https +48, asserts +56, pico2_w_release +24), pico_wiretft_angulo by +48. Flash used (.bin) against the branch's records: release 1,022,396 -> 1,022,660 (+264, 1,034,692), test +216 (1,027,092, 1,039,124), test_https +256 (1,032,140), asserts +264 (1,024,964, 1,036,996), pico2_w_release +248 (998,776, 998,800), pico_wire +568 (592,116, 604,152; the 1.3.1 send path, as main measured), pico_wiretft +896 (774,524, 786,560), pico_wiretft_angulo +904 (819,120, 831,152; RAM 116,744 -> 116,812 B). pico_w_alpha and pico_w_air are byte for byte the branch's previous images (sha256 against stages/s8a_c5): the clock code is compiled out where there is no panel. 2026-10-21 sync2/ui-drivers-main: main 90e82d1 (#279, #280) merged into d9cdda4. #279 touches no firmware; #280 changes how an injected touch ends, in handleTouch( ), loopCore1( ) and mapTouchPoint( ). All ten images rebuilt from scratch with the default core, against the same ten of d9cdda4 built the same way. Flash used: release +48 B (used 1,022,700, .bin 1,034,732), test +56 B (used 1,027,132, .bin 1,039,164), test_https +48 B (used 1,032,180, .bin 1,044,212), asserts +56 B (used 1,025,004, .bin 1,037,036), pico2_w_release +16 B (used 998,784, .bin 998,808), wiretft +56 B (used 774,572, .bin 786,608); alpha, air, wire, wiretft_angulo byte for byte the same. On main the same change was release +40, test +40, test_https +40, asserts +40, pico2_w_release +24 and wiretft +48; the branch compiles handleTouch( ) and loopCore1( ) differently around it. In the normalized disassembly (tools/elf_funcoes.py) each image that changed differs from d9cdda4 in loopCore1( ), mapTouchPoint( ) and handleTouch( ) only, plus _fini( ) where its trailing alignment halfword came or went, and on the Pico 2 W the binary-info header word that holds an address. 2026-10-21 S9 C2 (feat/ui-drivers-s9): the PIN keypad's touch becomes core actions, compiled into the Angulo's glue as well as into the original's handleTouchPanelV24( ), which the Angulo image links until S10b: pico_wiretft_angulo 821,608 B (+600 B) and a .bin of 833,640 B; the nine product images byte for byte unchanged. 2026-10-21 S9 C3 (feat/ui-drivers-s9): the sensor menu's permission table moves into its core action, so its closure, an IPA-SRA clone now, builds the table itself and handleTouchPanelV24( ) no longer does: the six TFT product images 8 B smaller (release, test, test_https, asserts, pico2_w_release, pico_wiretft; budgets lowered with them), pico_wiretft_angulo 16 B larger (821,624 B, .bin 833,656 B); same calls in the same order and the same conditional branches in every changed function (tools_s9/calls_check.py). 2026-10-21 S9 C4 (feat/ui-drivers-s9): the account editor's permission table moves into its core action, which the handler's inlined closure now builds in the row's own path, so handleTouchPanelV24( )'s frame is 8 B smaller: five TFT product images 8 B smaller (release, test, test_https, asserts, pico_wiretft; budgets lowered with them), pico2_w_release the same size with different bytes, pico_wiretft_angulo 8 B larger (821,632 B, .bin 833,664 B); same calls in the same order and the same conditional branches in every changed product function (tools_s9/calls_check.py, which counts the RP2350's cbz / cbnz as the beq / bne they stand for since this step). 2026-10-21 feat/ui-drivers-s9, S9 C6 (the password keyboard's typing as two core actions): the nine product images are byte-equal to C5's and C4's; only pico_wiretft_angulo changes, 8 B smaller (its digitar( ) lambda calls actPwdAppend( ) instead of storing twice): measured 821624 B and .bin 833656 B, the numbers of C3, so its budget goes down with it and its OTA row too. 2026-10-18 fix/safe-reboot-writes-pending-log: safeReboot( ) writes the log records it holds in RAM (flushPendingIfAny( ) inside the pre-reboot hook's guard, in safeReboot( ) and in the RP2350's safeRebootFlashUpdate( )). Four images grow 8 B, the call: pico_w_test (1,027,204, .bin 1,039,236), pico_w_asserts (1,025,076, .bin 1,037,108), pico2_w_release (998,880, .bin 998,904) and pico_wire (592,124, .bin 604,160). pico_w_release, pico_w_test_https, pico_w_alpha, pico_w_air and pico_wiretft come out the size of their records, flash used and .bin, their alignment absorbing it; RAM unchanged. 2026-10-21 fix/radio-answers-by-return-code (#266, #282): a reconnect scan that started quickly and never finished asks the chip one read (radioRead( ), an antenna-setting ioctl) and goes by the driver's return code, a quick start no longer ends a silent streak, and the minute's RSSI sample and getRssi( ) read cyw43_wifi_get_rssi( )'s return code (readRssi( )) instead of the variable WiFi.RSSI( ) leaves unwritten. On main 7e3f29a6 (#281 in), the seven radio images grow: release +160 B (1,022,932, .bin 1,034,964), test +176 (1,027,380, .bin 1,039,412, 731 B of OTA headroom left), test_https +176 (1,032,428), asserts +160 (1,025,236, .bin 1,037,268), alpha +168 (976,180) and air +168 (1,019,512), both .bin the same, pico2_w_release +184 (999,064, .bin 999,088). The two wired images build no radio unit and are byte for byte main's (pico_wire sha256 f464769b, pico_wiretft e781ea41); RAM unchanged on all nine. 2026-10-21 sync3/ui-drivers-main: feat/ui-drivers 154ff60c takes main 2f4bc3d3 (#281, #282; the merge commit that carries this note). All ten images rebuilt from scratch on the merged tree, against ~/.platformio, and 154ff60c rebuilt the same way for the other side (flash used and .bin, before -> after): pico_w_release 1,022,684 -> 1,022,852 B used (+168), .bin 1,034,716 -> 1,034,884 (+168); pico_w_test 1,027,116 -> 1,027,292 B used (+176), .bin 1,039,148 -> 1,039,324 (+176); pico_w_test_https 1,032,164 -> 1,032,348 B used (+184), .bin 1,044,196 -> 1,044,380 (+184); pico_w_asserts 1,024,988 -> 1,025,148 B used (+160), .bin 1,037,020 -> 1,037,180 (+160); pico_w_alpha 976,012 -> 976,180 B used (+168), .bin 990,380 -> 990,380 (+0); pico_w_air 1,019,344 -> 1,019,512 B used (+168), .bin 1,035,340 -> 1,035,340 (+0); pico2_w_release 998,776 -> 998,968 B used (+192), .bin 998,800 -> 998,992 (+192); pico_wire 592,116 -> 592,124 B used (+8), .bin 604,152 -> 604,160 (+8); pico_wiretft 774,556 -> 774,564 B used (+8), .bin 786,592 -> 786,600 (+8); pico_wiretft_angulo 821,624 -> 821,632 B used (+8), .bin 833,656 -> 833,664 (+8). Every record above is the merged tree's measurement, not the sum of the two sides' text; the two sentences just above this one (2026-10-18 and 2026-10-21) are main's own, with main's sizes, on main's trees. Host only: nothing of this merge has run on a board. 2026-10-21 feat/ui-drivers-s10a C2: the limits editor's step in one action for both drivers (form Fb), the one step of S10a that changes the products; the three LCD images are byte for byte C0's and the seven TFT images were built here by tools_s10a/verify_step.sh: pico_w_release 1,022,852 -> 1,022,940 B used (+88), .bin 1,034,884 -> 1,034,972 (+88); pico_w_test 1,027,292 -> 1,027,388 B used (+96), .bin 1,039,324 -> 1,039,420 (+96); pico_w_test_https 1,032,348 -> 1,032,436 B used (+88), .bin 1,044,380 -> 1,044,468 (+88); pico_w_asserts 1,025,148 -> 1,025,244 B used (+96), .bin 1,037,180 -> 1,037,276 (+96); pico2_w_release 998,968 -> 998,960 B used (-8), .bin 998,992 -> 998,984 (-8); pico_wiretft 774,564 -> 774,660 B used (+96), .bin 786,600 -> 786,696 (+96); pico_wiretft_angulo 821,632 -> 821,688 B used (+56), .bin 833,664 -> 833,720 (+56). Host only: nothing of this change has run on a board. 2026-10-21 fix/language-pack-log-records (#283): the language pack's records get their own tag (TAG_I18N, 13) and codes (620-623), the console text of the four carries only its detail, and the tags passed to the LOG_ macros are checked at compile time. On main 2f4bc3d3, the seven images that build the pack loader shrink: release -24 B (1,022,908, .bin 1,034,940), test -32 (1,027,348, .bin 1,039,380, 763 B of OTA headroom left), test_https -24 (1,032,404), asserts -24 (1,025,212, .bin 1,037,244), air -32 (1,019,480, the .bin the same), pico2_w_release -48 (999,016, .bin 999,040), wiretft -24 (774,612, .bin 786,648); their budgets come down. On release: findAndLoadLangFile( ) -92 (its two String concatenations), translateCodeEn( ) +52 (the four new cases), tagStringToId( ) and loadLangFile( ) +4 each, .rodata +8 (the four names against the shorter console texts); the name table moved from tagIdToString( )'s function-local static to LOG_TAG_NAMES, 64 B either way. The two images that do not build the loader grow, raising their budgets: pico_wire +184 (592,308, .bin 604,344: translateCodeEn( ) +52 and tagStringToId( ) +4 in .text, the four names +128 in .rodata) and alpha +56 (976,236, the .bin the same: .text +56, its .rodata absorbing the names in the padding before BTstack's sector-aligned TLV). RAM unchanged on all nine. The check itself costs nothing: the same tree with no static_assert, tagStringToId( ) inline and the old function-local table builds byte-identical images (release sha256 e971dfe8, air e5935de2, wiretft a01a20b2). 2026-10-21 sync4/ui-drivers-main: feat/ui-drivers edc95d6d takes main 25e169df (#284, the language pack's log records; the merge commit that carries this note). All ten images rebuilt from scratch on the merged tree, against ~/.platformio, and edc95d6d rebuilt the same way for the other side (flash used and .bin, before -> after): pico_w_release 1,022,940 -> 1,022,916 B used (-24), .bin 1,034,972 -> 1,034,948 (-24); pico_w_test 1,027,388 -> 1,027,356 B used (-32), .bin 1,039,420 -> 1,039,388 (-32); pico_w_test_https 1,032,436 -> 1,032,412 B used (-24), .bin 1,044,468 -> 1,044,444 (-24); pico_w_asserts 1,025,244 -> 1,025,220 B used (-24), .bin 1,037,276 -> 1,037,252 (-24); pico_w_alpha 976,180 -> 976,236 B used (+56), .bin 990,380 -> 990,380 (+0); pico_w_air 1,019,512 -> 1,019,480 B used (-32), .bin 1,035,340 -> 1,035,340 (+0); pico2_w_release 998,968 -> 998,920 B used (-48), .bin 998,992 -> 998,944 (-48); pico_wire 592,124 -> 592,308 B used (+184), .bin 604,160 -> 604,344 (+184); pico_wiretft 774,660 -> 774,636 B used (-24), .bin 786,696 -> 786,672 (-24); pico_wiretft_angulo 821,432 -> 821,400 B used (-32), .bin 833,464 -> 833,432 (-32). Every record above is the merged tree's measurement, not the sum of the two sides' text; the sentence just above this one is main's own, with main's sizes, on main's trees. Host only: nothing of this merge has run on a board. 2026-10-21 s10b/c0b: pico2_w_angulo joins, the Angulo on the Pico 2 W (pico2_w_release with ui = angulo, publish = false; like its parent it does not trust the bench root). The ten images and the new one rebuilt from scratch on feat/ui-drivers-s10b 50828bdf plus this change, against ~/.platformio: pico2_w_angulo 1,041,276 B used, .bin 1,041,296, RAM 132,564 B (budget = used; the slot ceiling 1,568,768 B as on pico2_w_release); against pico2_w_release on the same tree (998,920 B used, .bin 998,944) it is +42,356 B used and +42,352 B of .bin. The ten others are byte for byte what sync4 and C0 recorded (the nine sha256 and the Wire Angulo 44e1220f), so no other record moves. Host only: nothing of this change has run on a board. 2026-10-21 s10b/c2: pico_wiretft_angulo 821,400 -> 821,440 B used (+40), .bin 833,432 -> 833,472 (+40); pico2_w_angulo 1,041,276 -> 1,041,308 B used (+32), .bin 1,041,296 -> 1,041,328 (+32) (built from scratch with tools/build9.sh after the update screen's case called actOtaDismiss( ) itself and uaToque( ) answered true on every path). 2026-10-21 s10b/c3: pico_wiretft_angulo 821,440 -> 808,848 B used (-12,592), .bin 833,472 -> 820,880 (-12,592); pico2_w_angulo 1,041,308 -> 1,029,876 B used (-11,432), .bin 1,041,328 -> 1,029,896 (-11,432) (built from scratch with tools/build9.sh after the original's touch chain moved out of the Angulo's link (about 12.5 kB of the chain and handleTouchPanelV24( ) and what only they reached)). 2026-10-21 s10b/c4a: pico_wiretft_angulo 808,848 -> 808,728 B used (-120), .bin 820,880 -> 820,760 (-120); pico2_w_angulo 1,029,876 -> 1,029,748 B used (-128), .bin 1,029,896 -> 1,029,768 (-128) (built from scratch with tools/build9.sh after Themes.cpp left the Angulo's link and the Angulo defined the theme catalogue in ui/angulo/Temas.cpp (its three rows; the original's palettes, the .thm parser and its 800 B of custom themes leave)). 2026-10-21 s10b/c4b: pico_wiretft_angulo 808,728 -> 808,904 B used (+176), .bin 820,760 -> 820,936 (+176); pico2_w_angulo 1,029,748 -> 1,029,948 B used (+200), .bin 1,029,768 -> 1,029,968 (+200) (the .thm upload is refused at UPLOAD_FILE_START where SIMUT_UI_CUSTOM_THEMES is 0 (the Angulo images): WebManager::handleUploadData( ) and the new text; the nine compile it out). 2026-10-21 s10b/c5b: pico_wiretft_angulo 808,904 -> 808,896 B used (-8), .bin 820,936 -> 820,928 (-8); pico2_w_angulo 1,029,948 -> 1,029,932 B used (-16), .bin 1,029,968 -> 1,029,952 (-16) (the _uaPintando guards of fastFillRect, fastClearScreen and blitCanvas, the member and its two writes leave: the Angulo image does not link the three primitives; only the constructor, uaPintar and the offset of the member after the removed one change). S10c A'4, 2026-10-22: pico_wiretft_angulo 808,896 -> 808,536 B used (-360), .bin 820,928 -> 820,568 (-360); pico2_w_angulo 1,029,932 -> 1,029,556 B used (-376), .bin 1,029,952 -> 1,029,576 (-376) (a .thm upload is stored, not refused: C4b's refusal block and the two theme rescans leave the Angulo images, three strings go (.thm, Custom themes rescanned, the refusal text); only WebManager::handleUploadData and WebManager::handleDelete change in code, the rest is layout). 2026-10-22 s10c/a2: pico_wiretft_angulo 808,536 -> 809,408 B used (+872), .bin 820,568 -> 821,440 (+872); pico2_w_angulo 1,029,556 -> 1,030,524 B used (+968), .bin 1,029,576 -> 1,030,544 (+968) (A'2 puts the palette's 24 generated tables and the mode and accent functions into the two Angulo images). 2026-10-22 s10c/a3: pico_wiretft_angulo 809,408 -> 811,664 B used (+2,256), .bin 821,440 -> 823,696 (+2,256); pico2_w_angulo 1,030,524 -> 1,032,796 B used (+2,272), .bin 1,030,544 -> 1,032,816 (+2,272) (A'3's Tema screen (segments, swatches, button), the preview and the catalogue of 18 enter the two Angulo images). 2026-10-22 s10c/a3: pico_wiretft_angulo 811,664 -> 811,664 B used (+0), .bin 823,696 -> 823,696 (+0); pico2_w_angulo 1,032,796 -> 1,032,796 B used (+0), .bin 1,032,816 -> 1,032,816 (+0) (A'3's Tema screen (segments, swatches, button), the preview and the catalogue of 18 enter the two Angulo images). 2026-10-22 s10c/a2b: pico_wiretft_angulo 811,664 -> 811,300 B used (-364), .bin 823,696 -> 823,336 (-360); pico2_w_angulo 1,032,796 -> 1,032,400 B used (-396), .bin 1,032,816 -> 1,032,424 (-392) (the Angulo's palette is built by the compiler (constexpr tables from the generator) and the static initialiser is gone). 2026-10-22 s10c/b2b: pico_wiretft_angulo 811,300 -> 812,244 B used (+944), .bin 823,336 -> 824,280 (+944); pico2_w_angulo 1,032,400 -> 1,033,304 B used (+904), .bin 1,032,424 -> 1,033,328 (+904) (the Angulo loads its texts from the language pack at boot: Textos.cpp joins the two images, the hook, the scanner's @ANGULO entry, two words in ActiveLang). 2026-10-22 S10c B3 (the Painel and Base screens look their texts up in the language pack): pico_wiretft_angulo 812,244 -> 812,972 B used (+728), .bin 824,280 -> 825,008 (+728); pico2_w_angulo 1,033,304 -> 1,033,976 B used (+672), .bin 1,033,328 -> 1,034,000 (+672) (the English literals and the lookups replace the Portuguese literals; the nine products are byte-identical). 2026-10-22 feat/ui-drivers-s10c B3-b: pico_wiretft_angulo 812,972 -> 812,924 B used (-48), .bin 825,008 -> 824,960 (-48); pico2_w_angulo 1,033,976 -> 1,033,920 B used (-56), .bin 1,034,000 -> 1,033,944 (-56) (the kit's back action is one function, acaoVoltar( ), which the three screens that rebuilt it with the literal now call; static RAM unchanged, 115,912 and 131,572 B). 2026-10-22 feat/ui-drivers-s10c B3-c: pico_wiretft_angulo 812,924 -> 812,940 B used (+16), .bin 824,960 -> 824,976 (+16); pico2_w_angulo 1,033,920 -> 1,033,936 B used (+16), .bin 1,033,944 -> 1,033,960 (+16) (the pager's count is one kit function, contagemDoPaginador( ), which paginador( ) draws and the charts' page name is placed after; static RAM unchanged, 115,912 and 131,572 B). 2026-10-22 feat/ui-drivers-s10c B4a: pico_wiretft_angulo 812,940 -> 813,588 B used (+648), .bin 824,976 -> 825,624 (+648); pico2_w_angulo 1,033,936 -> 1,034,544 B used (+608), .bin 1,033,960 -> 1,034,568 (+608) (the Menus screens' texts are UA_TXT( ) lookups, the English in the source and the Portuguese or the Spanish in the packs; static RAM unchanged, 115,912 and 131,572 B). 2026-10-22 feat/ui-drivers-s10c B4b-0: pico_w_release 1,022,916 -> 1,022,900 B used (-16), .bin 1,034,948 -> 1,034,932 (-16); pico_w_test 1,027,356 -> 1,027,348 B used (-8), .bin 1,039,388 -> 1,039,380 (-8); pico_w_asserts 1,025,220 -> 1,025,204 B used (-16), .bin 1,037,252 -> 1,037,236 (-16); pico2_w_release 998,920 -> 998,928 B used (+8), .bin 998,944 -> 998,952 (+8); pico_wire 592,308 -> 592,268 B used (-40), .bin 604,344 -> 604,304 (-40); pico_wiretft 774,636 -> 774,596 B used (-40), .bin 786,672 -> 786,632 (-40); pico_wiretft_angulo 813,588 -> 813,548 B used (-40), .bin 825,624 -> 825,584 (-40); pico2_w_angulo 1,034,544 -> 1,034,552 B used (+8), .bin 1,034,568 -> 1,034,576 (+8) (LANG_FILE_MAX 49,152 -> 65,536 and lang.js's comment of the viewer's cap: the comment re-stamps every web page (the ?v= hash of the raw lang.js), so each page's gzip moves by a few bytes; the normalized pages are identical, and loadLangFile( )'s constant is the only code change). 2026-10-22 feat/ui-drivers-s10c B4b: pico_wiretft_angulo 813,548 -> 814,028 B used (+480), .bin 825,584 -> 826,064 (+480); pico2_w_angulo 1,034,552 -> 1,035,008 B used (+456), .bin 1,034,576 -> 1,035,032 (+456) (the Ajustes screens' texts are UA_TXT( ) lookups, the English in the source and the Portuguese or the Spanish in the packs, and the footer's boxes keep the labels they measure; static RAM unchanged, 115,912 and 131,572 B). 2026-10-22 feat/ui-drivers-s10c B5a: pico_wiretft_angulo 814,028 -> 814,356 B used (+328), .bin 826,064 -> 826,392 (+328); pico2_w_angulo 1,035,008 -> 1,035,392 B used (+384), .bin 1,035,032 -> 1,035,416 (+384) (the Gráficos screens' texts are UA_TXT( ) lookups, the English in the source and the Portuguese or the Spanish in the packs, n1( ) and kb( ) write the language's decimal separator; static RAM unchanged, 115912 and 131572 B). 2026-10-23 feat/ui-drivers-s10c B5a-b: pico_wiretft_angulo 814,356 -> 814,364 B used (+8), .bin 826,392 -> 826,400 (+8); pico2_w_angulo 1,035,392 -> 1,035,400 B used (+8), .bin 1,035,416 -> 1,035,424 (+8) (ua::separadorDecimal( ) takes only ',' or '.' from a pack; static RAM unchanged). 2026-10-23 feat/ui-drivers-s10c B5b: pico_wiretft_angulo 814,364 -> 814,700 B used (+336), .bin 826,400 -> 826,736 (+336); pico2_w_angulo 1,035,400 -> 1,035,592 B used (+192), .bin 1,035,424 -> 1,035,616 (+192) (the Identidade screens' texts are UA_TXT( ) lookups, the English in the source and the Portuguese or the Spanish in the packs, the identity bar and footers keep the labels their boxes were measured from, the Painel's missing-value placeholder takes the language's separator; static RAM unchanged, 115912 and 131572 B). 2026-10-22 fix/alpha-console-event-names: the alpha's two log lookups, logcodeLookup( ) and trlLookup( ) in DisplayManager_Alpha.cpp, answer nullptr where they answered \"\", which LogManager took as the translation. Each became `movs r0, #0; bx lr`, 4 B, where it loaded a literal, 8 B: pico_w_alpha -8 B (976,228, the .bin the same) and pico_wire -8 B (592,300, .bin 604,336), the two images that build that file. The seven others are byte for byte main's (25e169d), compared by sha256. RAM unchanged on all nine. 2026-10-22 chore/https-test-image-room: pico_w_test_https serves the license page from LittleFS too (custom_fs_pages ALARMS_PAGE, FILE_PAGE, TEL_PAGE and LICENSE_PAGE), the maintainer's option (a) for feat/ui-drivers' S10b-2, which grows this image about 160 B where its FLASH region had 44 B left (the last LOAD ended at 0x100FEFD4, readelf -l): -4,256 B (1,028,148, the .bin 1,040,180 and not recorded, ota_exempt), the region's room 4,300 B (the last LOAD ends at 0x100FDF34); its budget comes down. The license page is 4,260 B of gzip in this env. The other eight are byte for byte main's (a8ccf4c3), compared by sha256. 2026-10-23 sync5/ui-drivers-main: feat/ui-drivers ecb527e3 takes main 2075b84e (#285 to #288: the savings floor past the BTstack fill, the console and syslog event names on alpha and wire, the docs of the 22/10 bench, and pico_w_test_https's licence page on LittleFS; the merge commit that carries this note). All eleven images rebuilt from scratch on the merged tree, against ~/.platformio, and ecb527e3 rebuilt the same way for the other side (flash used and .bin, before -> after): pico_w_release 1,022,900 -> 1,022,900 B used (+0), .bin 1,034,932 -> 1,034,932 (+0); pico_w_test 1,027,348 -> 1,027,348 B used (+0), .bin 1,039,380 -> 1,039,380 (+0); pico_w_test_https 1,032,412 -> 1,028,148 B used (-4264), .bin 1,044,444 -> 1,040,180 (-4264); pico_w_asserts 1,025,204 -> 1,025,204 B used (+0), .bin 1,037,236 -> 1,037,236 (+0); pico_w_alpha 976,236 -> 976,228 B used (-8), .bin 990,380 -> 990,380 (+0); pico_w_air 1,019,480 -> 1,019,480 B used (+0), .bin 1,035,340 -> 1,035,340 (+0); pico2_w_release 998,928 -> 998,928 B used (+0), .bin 998,952 -> 998,952 (+0); pico_wire 592,268 -> 592,260 B used (-8), .bin 604,304 -> 604,296 (-8); pico_wiretft 774,596 -> 774,596 B used (+0), .bin 786,632 -> 786,632 (+0); pico_wiretft_angulo 814,716 -> 814,716 B used (+0), .bin 826,752 -> 826,752 (+0); pico2_w_angulo 1,035,608 -> 1,035,608 B used (+0), .bin 1,035,632 -> 1,035,632 (+0). Every record above is the merged tree's measurement, not the sum of the two sides' text; the sentence just above this one is main's own, with main's sizes, on main's trees. Host only: nothing of this merge has run on a board. S10c D5 (2026-10-23): pico_w_release 1,022,900 -> 1,022,908 B used (+8), .bin 1,034,932 -> 1,034,940 (+8); pico_w_test_https 1,028,148 -> 1,028,156 B used (+8), exempt from the OTA ceiling so no .bin recorded; pico2_w_release 998,928 -> 998,952 B used (+24), .bin 998,952 -> 998,976 (+24) (the alarm screen's three draw functions moved to DisplayManager_AlarmUi.cpp, so redrawAlarmFlash( ) reads _lastRenderedState.topSlotIdx again after isSlotAlarming( ) and isSlotErrAlarming( ), which it can no longer see only read (141 -> 143 instructions on the RP2040, 104 -> 110 on the Pico 2 W, where slotAlarmBg( ) also saves this and the slot in r4 and r5 across its calls); host only, nothing ran on a board). S10c D3a: pico_w_release 1,022,908 -> 1,022,916 B used (+8), .bin 1,034,940 -> 1,034,948 (+8); pico_w_test 1,027,348 -> 1,027,364 B used (+16), .bin 1,039,380 -> 1,039,396 (+16); pico_w_test_https 1,028,156 -> 1,028,164 B used (+8), exempt from the OTA ceiling so no .bin recorded; pico_w_asserts 1,025,204 -> 1,025,220 B used (+16), .bin 1,037,236 -> 1,037,252 (+16); pico2_w_release 998,952 -> 998,960 B used (+8), .bin 998,976 -> 998,984 (+8); pico_wiretft 774,596 -> 774,612 B used (+16), .bin 786,632 -> 786,648 (+16); pico_wiretft_angulo 814,716 -> 814,644 B used (-72), .bin 826,752 -> 826,680 (-72); pico2_w_angulo 1,035,608 -> 1,035,528 B used (-80), .bin 1,035,632 -> 1,035,552 (-80) (the seven calibration draw-cache stores moved into resetTouchCalCaches( ) behind UiChange::TOUCH_CAL_ENTER). S10c D3b: pico_wiretft_angulo 814,644 -> 814,636 B used (-8), .bin 826,680 -> 826,672 (-8); pico2_w_angulo 1,035,528 -> 1,035,536 B used (+8), .bin 1,035,552 -> 1,035,560 (+8) (showSettingsDisplayOffset( ) sets the original's offset sentinels behind UiChange::DISPLAY_OFFSET_ENTER). S10c D6a: pico_w_release 1,022,916 -> 1,023,084 B used (+168), .bin 1,034,948 -> 1,035,116 (+168); pico_w_test 1,027,364 -> 1,027,524 B used (+160), .bin 1,039,396 -> 1,039,556 (+160); pico_w_test_https 1,028,164 -> 1,028,324 B used (+160); pico_w_asserts 1,025,220 -> 1,025,380 B used (+160), .bin 1,037,252 -> 1,037,412 (+160); pico2_w_release 998,960 -> 999,072 B used (+112), .bin 998,984 -> 999,096 (+112); pico_wiretft 774,612 -> 774,772 B used (+160), .bin 786,648 -> 786,808 (+160); pico_wiretft_angulo 814,636 -> 814,700 B used (+64), .bin 826,672 -> 826,736 (+64); pico2_w_angulo 1,035,536 -> 1,035,592 B used (+56), .bin 1,035,560 -> 1,035,616 (+56) (the original's user and PIN screens move to DisplayManager_UsersUi.cpp, and GCC inlines PinKb::orderedFace, clampPinPolicy and maxLenFor once their callers are in two units). 2026-10-28 S10c D7: pico2_w_angulo 1,035,592 -> 1,035,992 B used (+400), .bin 1,035,616 -> 1,036,016 (+400) (the setup-network confirmation's frame and touch, three texts, one #if SIMUT_NET_WIFI). 2026-10-24 fix/lang-english-way-back: after a boot in English the Idioma screen lists the pack in /lang without loading it (display/LangChoice.h, listLangFile( ), the choice saved at once and loaded at the next boot): pico_w_release +616 B (1,023,524, the .bin 1,035,556), pico_w_test +624 (1,027,972, the .bin 1,040,004: 139 B of OTA headroom left of 763), pico_w_asserts +608 (1,025,820, the .bin 1,037,852), pico2_w_release +656 (999,672, the .bin 999,696), pico_wiretft +624 (775,236, the .bin 787,272), and pico_w_test_https +624 (1,028,772, the .bin 1,040,804 and not recorded, ota_exempt), which leaves its FLASH region 3,676 B of room (the last LOAD ends at 0x100FE1A4, readelf -l). pico_w_air -16 B (1,019,464, the .bin the same size, 1,035,340): its sha256 differs, because loadLangFile( ) and findAndLoadLangFile( ) are compiled differently (the one rule for a usable pack, and the directory pass the list shares); pico_w_alpha and pico_wire are byte for byte main's (be56633d), compared by sha256. RAM unchanged on all nine: a listed pack is told by the path kept in _activeLang, not by a new variable. 2026-10-28 fix/panel-theme-lang-permission on d2c366b3 (PR #294's head): the guard arm that refuses EVT_APPLY_THEME and EVT_APPLY_LANG without PERM_SYS_CONFIG sits in the dispatch chain's TFT region, so only the six panel images compile it: release +64 B (1,023,588, .bin 1,035,620), test +64 (1,028,036, .bin 1,040,068; OTA room 139 -> 75 B), test_https +72 (1,028,844), asserts +72 (1,025,892, .bin 1,037,924), pico2_w_release +24 (999,696, .bin 999,720), pico_wiretft +64 (775,300, .bin 787,336); pico_w_alpha, pico_w_air and pico_wire are byte for byte d2c366b3's, compared by sha256 (b5d7d079, 4291184e, 3438b89a). RAM unchanged on all nine. 2026-10-28 sync 6 (C1): pico_w_release 1,023,084 -> 1,023,764 B used (+680), .bin 1,035,116 -> 1,035,796 (+680); pico_w_test 1,027,524 -> 1,028,212 B used (+688), .bin 1,039,556 -> 1,040,244 (+688); pico_w_asserts 1,025,380 -> 1,026,060 B used (+680), .bin 1,037,412 -> 1,038,092 (+680); pico_w_alpha 976,228 -> 976,228 B used (+0), .bin 990,380 -> 990,380 (+0); pico_w_air 1,019,480 -> 1,019,464 B used (-16), .bin 1,035,340 -> 1,035,340 (+0); pico2_w_release 999,072 -> 999,744 B used (+672), .bin 999,096 -> 999,768 (+672); pico_wire 592,260 -> 592,260 B used (+0), .bin 604,296 -> 604,296 (+0); pico_wiretft 774,772 -> 775,452 B used (+680), .bin 786,808 -> 787,488 (+680); pico_wiretft_angulo 814,700 -> 815,300 B used (+600), .bin 826,736 -> 827,336 (+600); pico2_w_angulo 1,035,992 -> 1,036,576 B used (+584), .bin 1,036,016 -> 1,036,600 (+584); pico_w_test_https 1,028,324 -> 1,029,020 B used (+696), no .bin recorded (ota_exempt) (main's #289-#295 merged into the branch: the English boot lists the pack, the Idioma choice waits for a boot, the guard arm; the changed functions are main's own pair, decoded against it). sync 6 C3: pico_wiretft_angulo 815,300 -> 815,356 B used (+56), .bin 827,336 -> 827,392 (+56); pico2_w_angulo 1,036,576 -> 1,036,632 B used (+56), .bin 1,036,600 -> 1,036,656 (+56) (the Angulo's Idioma screen lists the pack an English boot found (#294)). 2026-10-28 chore/test-image-license-fs: pico_w_test serves the license page from LittleFS too (custom_fs_pages LICENSE_PAGE), the maintainer's \"Abrir espaco na pico_w_test\": it signed 1,040,309 B against the OTA ceiling of 1,040,384 B, 75 B of room on main 9f4da77b, and the UI-drivers branch after its sync with main was 101 B over. -3,624 B (flash used 1,024,412, .bin 1,036,444, signed 1,036,685, 3,699 B of room; RAM unchanged at 109,212 B), not the 4,260 B of the page's gzip: no page of this image was served from the file before, so serveProtectedFsPage( ) (428 B; handleLicense( ) 8 B smaller) and the \"Page asset missing\" notice with the file name (212 B of strings in .rodata, 4 B of .text outside any symbol) are linked now; .text +424, .rodata -4,048. Its budget and \"bin\" come down. The other eight are byte for byte main's (9f4da77b), compared by sha256. sync 6b: pico_w_test 1,028,212 -> 1,024,580 B used (-3,632), .bin 1,040,244 -> 1,036,612 (-3,632) (main's #297 merged: pico_w_test serves its licence page from LittleFS (LICENSE_PAGE_GZ out, serveProtectedFsPage( ) in, handleLicense( ) 24 -> 16 B)). 2026-10-28 feat/web-theme-names on 99e9001b (main after #297): GET /api/themes sends each theme's code and the dashboard translates the stock names by it (formatThemeRow( ) in src/ThemeApiRow.h, themeOptions( ) in the TFT block of WebUI.h, the row buffer 96 B; each option carries data-en, the device's English, so that applyLang( ) restores it when /api/lang answered first). The six panel images grow: release +160 B (1,023,748, .bin 1,035,780; OTA room 4,523 -> 4,363), test +160 (1,024,572, .bin 1,036,604; 3,699 -> 3,539), test_https +168 (1,029,012; its .bin 1,041,044 is not recorded, ota_exempt), asserts +152 (1,026,044, .bin 1,038,076; 2,219 -> 2,067), pico2_w_release +168 (999,864, .bin 999,888) and pico_wiretft +160 (775,460, .bin 787,496). The three images with no panel do not link handleApiThemes( ), but its string literals sit in WebManager_Api.o's merged string pool, which the linker keeps whole: pico_wire +16 B (592,316, .bin 604,352), and pico_w_alpha (976,228, .bin 990,380) and pico_w_air (1,019,464, .bin 1,035,340) keep their sizes with another sha256, every symbol after the first changed string 16 B further on inside a .rodata of 434,176 and 454,656 B (multiples of 4,096), their page bytes identical (the page change of data-en is inside the TFT block and leaves them as they were). RAM unchanged on all nine. The same handler under #if SIMUT_DISPLAY_TFT, tried in a scratch tree, gave pico_wire 592,268 (-32 B against the budget it had) and was not taken: it would add an #if site to WebManager_Api.cpp. 2026-10-28 sync 7 (feat/ui-drivers + main c9028947 = #298, from scratch against the baseline 290c2662 built the same way): pico_w_release 1,023,764 -> 1,023,924 B used (+160), .bin 1,035,796 -> 1,035,956 (+160); pico_w_test 1,024,580 -> 1,024,740 B used (+160), .bin 1,036,612 -> 1,036,772 (+160); pico_w_asserts 1,026,060 -> 1,026,228 B used (+168), .bin 1,038,092 -> 1,038,260 (+168); pico_w_alpha 976,228 -> 976,228 B used (+0), .bin 990,380 -> 990,380 (+0); pico_w_air 1,019,464 -> 1,019,464 B used (+0), .bin 1,035,340 -> 1,035,340 (+0); pico2_w_release 999,744 -> 999,912 B used (+168), .bin 999,768 -> 999,936 (+168); pico_wire 592,260 -> 592,276 B used (+16), .bin 604,296 -> 604,312 (+16); pico_wiretft 775,452 -> 775,628 B used (+176), .bin 787,488 -> 787,664 (+176); pico_wiretft_angulo 815,356 -> 815,532 B used (+176), .bin 827,392 -> 827,568 (+176); pico2_w_angulo 1,036,632 -> 1,036,792 B used (+160), .bin 1,036,656 -> 1,036,816 (+160) (the theme-name page and route land on the branch: pico_w_release, pico_w_test, pico2_w_angulo +160, pico_w_asserts +168, pico2_w_release +168, pico_wiretft and pico_wiretft_angulo +176 (themeCleanCopy( ) +48 B, handleApiThemes( ) -20 B, the page's gzip +121..+129 B, the rows' strings, alignment), pico_wire +16 B (the strings), pico_w_alpha and pico_w_air the same size with other bytes; pico_w_test_https +168 B). 2026-10-28 sync 7, C2 (the Angulo's 18 theme names in English): pico_wiretft_angulo 815,532 -> 815,468 B used (-64), .bin 827,568 -> 827,504 (-64); pico2_w_angulo 1,036,792 -> 1,036,728 B used (-64), .bin 1,036,816 -> 1,036,752 (-64) (only the two Angulo images move, by the catalogue's strings in Temas.cpp (Portuguese names -> English); the nine products do not link it and are sha256-equal to the merge's). 2026-10-28 fix/web-upload-refused: doUpload( ) reads the answer of /api/upload, and a refusal gets a message of its own (fil_up_refused, English default in the page): FILE_PAGE_GZ 6,670 -> 6,779 B in pico_w_release, and +112 B measured and +112 B on the .bin in release, test, asserts, pico2_w_release, pico_wire and pico_wiretft. alpha and air carry the new page (6,779 B) and stay the same size, the alignment absorbing it; test_https carries no Files page and is unchanged. All nine rebuilt from scratch on c9028947 plus this change; main's own pico_w_release rebuilt first, at the recorded 1,023,748 B and a .bin of 1,035,780 B. 2026-10-30 feat/ui-drivers sync 8: pico_w_release 1,023,924 -> 1,024,028 B used (+104), .bin 1,035,956 -> 1,036,060 (+104); pico_w_test 1,024,740 -> 1,024,844 B used (+104), .bin 1,036,772 -> 1,036,876 (+104); pico_w_asserts 1,026,228 -> 1,026,332 B used (+104), .bin 1,038,260 -> 1,038,364 (+104); pico_w_alpha 976,228 -> 976,228 B used (+0), .bin 990,380 -> 990,380 (+0); pico_w_air 1,019,464 -> 1,019,464 B used (+0), .bin 1,035,340 -> 1,035,340 (+0); pico2_w_release 999,912 -> 1,000,024 B used (+112), .bin 999,936 -> 1,000,048 (+112); pico_wire 592,276 -> 592,388 B used (+112), .bin 604,312 -> 604,424 (+112); pico_wiretft 775,628 -> 775,740 B used (+112), .bin 787,664 -> 787,776 (+112); pico_wiretft_angulo 815,468 -> 815,580 B used (+112), .bin 827,504 -> 827,616 (+112); pico2_w_angulo 1,036,728 -> 1,036,840 B used (+112), .bin 1,036,752 -> 1,036,864 (+112) (the merge of main cf97b955 (#299, the Files page says a refused upload was refused; #300, SIMUT_VERSION 2.12.0; #301 and #302 touch no firmware input) into 22d83de3, all eleven rebuilt from scratch through the pio shim, the baseline 22d83de3 rebuilt the same way first and sha256-equal to sync 7's in all eleven. The only symbols that move are WebUI_GZ::FILE_PAGE_GZ (+109..+112 B) and LOGIN_PAGE_GZ (+1 B), as in main's own pair c9028947 -> cf97b955; .rodata moves 104 B in pico_w_release, pico_w_test and pico_w_asserts and 112 B in the other five page images (main's pair: 112 B in its six), the 8-byte alignment of the section absorbing the 110 B of symbols differently. pico_w_alpha and pico_w_air carry the new page and keep their size, pico_w_test_https carries no Files page and is unchanged). 2026-10-30 s10c/c4: pico2_w_release 1,036,840 -> 1,037,864 B used (+1,024), .bin 1,036,864 -> 1,037,888 (+1,024); pico_wiretft_angulo 815,580 -> 816,660 B used (+1,080), .bin 827,616 -> 828,696 (+1,080) (the Angulo's fonts and icons on /license (the @IF ui_angulo block, in the two Angulo images only) and on its licence screen (a heading and three paragraphs, four UA_TXT texts); the nine images that do not carry the Angulo keep their sha256). 2026-10-30 s10c/c4 review (sources pinned): pico2_w_release 1,037,864 -> 1,037,832 B used (-32), .bin 1,037,888 -> 1,037,856 (-32); pico_wiretft_angulo 816,660 -> 816,612 B used (-48), .bin 828,696 -> 828,648 (-48) (the Angulo's licence sources are a local table appended all or none (anexarFontesDoAngulo), smaller than the four guarded appends it replaces). 2026-10-30 s10c/c4 review (page labels): pico2_w_release 1,037,832 -> 1,037,848 B used (+16), .bin 1,037,856 -> 1,037,872 (+16); pico2_w_original 1,000,024 -> 1,000,040 B used (+16), .bin 1,000,048 -> 1,000,064 (+16); pico_wire 592,388 -> 592,388 B used (+0), .bin 604,424 -> 604,424 (+0); pico_wiretft 775,740 -> 775,740 B used (+0), .bin 787,776 -> 787,776 (+0); pico_wiretft_angulo 816,612 -> 816,612 B used (+0), .bin 828,648 -> 828,648 (+0); pico_w_release 1,024,028 -> 1,024,060 B used (+32), .bin 1,036,060 -> 1,036,092 (+32); pico_w_test 1,024,844 -> 1,024,844 B used (+0), .bin 1,036,876 -> 1,036,876 (+0); pico_w_asserts 1,026,332 -> 1,026,364 B used (+32), .bin 1,038,364 -> 1,038,396 (+32); pico_w_alpha 976,228 -> 976,228 B used (+0), .bin 990,380 -> 990,380 (+0); pico_w_air 1,019,464 -> 1,019,464 B used (+0), .bin 1,035,340 -> 1,035,340 (+0) (the FreeSans Bold row of the /license page names its four radio images instead of the five-image label, so the licence blob is 13 to 27 B longer on the radio images (pico_w_test and pico_w_test_https keep that page in LittleFS, pico_w_test_https has no budget row to move, and the Wire images keep their bytes)). 2026-10-30 fix/angulo-wifi-status: pico2_w_release 1,037,848 -> 1,037,856 B used (+8), .bin 1,037,872 -> 1,037,880 (+8) (uaVmStatus maps the Wi-Fi levels: the Angulo's status bar on the Pico 2 W release). 2026-10-30 fix/angulo-wifi-status r2: pico2_w_release 1,037,856 -> 1,038,032 B used (+176), .bin 1,037,880 -> 1,038,056 (+176); pico_wiretft_angulo 816,612 -> 816,764 B used (+152), .bin 828,648 -> 828,800 (+152) (the status bar's Wi-Fi levels: three Lucide icons (wifi-zero, wifi-low, wifi-high) in the shared registry, the shifted icon names, the new ladder in uaVmStatus( )). 2026-10-30 fix/angulo-wifi-status r3: pico2_w_release 1,038,032 -> 1,038,080 B used (+48), .bin 1,038,056 -> 1,038,104 (+48); pico_wiretft_angulo 816,764 -> 816,804 B used (+40), .bin 828,800 -> 828,840 (+40) (the status bar's Wi-Fi levels as Lucide's signal bars (four icons in, wifi-zero/low/high out of the registry), the cable without link in the neutral grey). 2026-10-30 fix/angulo-wifi-status r4: pico2_w_release 1,038,080 -> 1,038,112 B used (+32), .bin 1,038,104 -> 1,038,136 (+32); pico_wiretft_angulo 816,804 -> 816,860 B used (+56), .bin 828,840 -> 828,896 (+56) (the cable without link has an icon of its own, Lucide's unplug (one icon more in the registry, the ids after UNPLUG shift by one)). 2026-10-30 (S10c C6b): pico2_w_release 1,038,112 -> 1,038,080 B used (-32), .bin 1,038,136 -> 1,038,104 (-32); pico_wiretft_angulo 816,860 -> 816,836 B used (-24), .bin 828,896 -> 828,872 (-24) (the web-busy card and the accounts-full message say what the code does). 2026-10-30 S11 (the original panel's 14 units moved to src/ui/original/): pico_w_asserts 1,026,364 -> 1,026,356 B used (-8), .bin 1,038,396 -> 1,038,388 (-8); pico_wiretft 775,740 -> 775,732 B used (-8), .bin 787,776 -> 787,768 (-8); pico2_w_original 1,000,040 -> 1,000,032 B used (-8), .bin 1,000,064 -> 1,000,056 (-8) (link order only: .text loses 8 B of alignment padding on these three, the other eight images keep their sizes). 2026-10-31 fix/alarm-limits-snapshot on 3ca4976f (F3 of the v2.13.0 Known list): the limits editor's Salvar posts a copy into a 68 B member of DisplayManager, a TFT-only member, so the eight TFT images grow and the three LCD images do not. Used and .bin move together: pico_w_release +104, pico_w_test +104, pico_w_test_https +104, pico_w_asserts +104, pico2_w_release +48, pico_wiretft +144, pico_wiretft_angulo +104, pico2_w_original +120; pico_w_alpha, pico_w_air and pico_wire were built from zero at both commits and their firmware.bin has the same sha256. Each budget and .bin is set to the new measure, and the OTA table (PLANO_STABLE 3.1) says the same. 2026-10-31 fix/keypad-route-angulo on 6981fd93 (F4 of the v2.13.0 Known list): on the Angulo, GET /api/keypad sends the faces, the state, the policy and targets and no grid and no popup, chosen by a UI hook that is a constant on each side (uiKeypadTargets( ): nullptr on the original, \"/api/ui\" on the Angulo), so the branch an image does not take is not in the image: the two Angulo images shrink, pico2_w_release -208 and pico_wiretft_angulo -216 (used and .bin move together), and the other nine have the same sha256 as at 6981fd93. Both budgets and .bin are lowered to the new measure, and the OTA table (PLANO_STABLE 3.1) says the same. 2026-11-04 fix/small-bugs on 2128deb1 (main 34fc44a9 merged in, then twelve small fixes; all eleven built from zero through the pio shim of 2026-10-28 under the shared build lock, -Werror holding; main's records equal main's images on this host as of the same day): used against main pico_w_release +448, pico_w_test +480, pico_w_test_https +472 (used only), pico_w_asserts +440, pico_w_alpha +472 (.bin 0: the growth went into the linker fill in front of the Bluetooth bank), pico_w_air +408 (.bin 0: the growth went into the linker fill in front of the Bluetooth bank), pico2_w_release +336, pico_wire +104, pico_wiretft +64, pico_wiretft_angulo −24, pico2_w_original +432. The twelve fixes and the share of each are in CHANGELOG.md (Unreleased). Each budget and .bin is set to the new measure, and the OTA table (PLANO_STABLE 3.1) says the same. 2026-11-04 fix/small-bugs, recorded again at 212ff4bd after the review of the branch; used against the values this replaces: pico_w_release +0, pico_w_test +0, pico_w_test_https +0, pico_w_asserts +0, pico_w_alpha +0, pico_w_air +0, pico2_w_release +208, pico_wire +0, pico_wiretft +0, pico_wiretft_angulo +216, pico2_w_original +0. 2026-11-04 feat/alpha-lcd-live on 10c314c4 (main d8871388 and #319 merged in at d92b08a2, then steps 5a, 5b and the font's notice; all eleven built from zero through the pio shim of 2026-10-28 under the shared build lock, -Werror holding): the alpha's live LCD view (the glass that Core 1 publishes, GET /api/lcd, the box on the dashboard and its font) is in the two images with the alpha's LCD and in no other: the other nine have the same sha256 and size as main's images (tip2 table of 2026-11-04, pico_w_release 48c3b497 to pico2_w_original 4b066838), and the decoded pages inside each firmware.bin carry the box and the font's notice in pico_w_alpha and pico_wire only. Against the records it replaces: pico_w_alpha 976,732 -> 981,916 B used (+5,184 B) and .bin 990,380 -> 994,476 B (+4,096 B: one 4 KB step, the rest went into the linker fill in front of the Bluetooth bank), RAM 125,788 B; pico_wire 592,532 -> 597,372 B used (+4,840 B) and .bin 604,568 -> 609,408 B (+4,840 B), RAM 104,592 B. Steps 1 to 3 (the glass's shadow, GET /api/lcd) are +1,088 B used on the alpha and +1,296 B on the Wire, and the RAM of +108 B on both (measured at d92b08a2: alpha 977,820 B used, .bin 990,380 B; Wire 593,828 B, .bin 605,864 B); step 5 is the rest, +4,096 B on the alpha and +3,544 B on the Wire, of which the notice is +312 B on the Wire and 0 B on the alpha. Each budget and .bin is set to the new measure, and the OTA table (PLANO_STABLE 3.1) says the same. 2026-11-05 feat/alpha-lcd-live, recorded again at 03d05e47 after the page's glyph table became ROM_A00 of rom_hd44780.py (the maintainer's decision of 2026-11-04: the live view shows the characters the real LCD shows; the earlier font and its /license notice are gone) and after the review's fixes (the age in its own element, only a frame resets the count of failures, the body read with a timeout); all eleven built from zero through the pio shim of 2026-10-28 under the shared build lock, -Werror holding, and the nine without the alpha's LCD are sha256- and size-equal to main's images (tip2 table of 2026-11-04, 9 of 9). Used against the values this replaces: pico_w_alpha +0 (981,916 B; .bin 994,476 B, +0: its linker fill absorbs the change) and pico_wire -160 (597,372 -> 597,212 B; .bin 609,408 -> 609,248 B). On the Wire the notice's removal is -312 B and the new table's data +24 B (597,084 B at 9c456124), and the review's fixes +128 B. Against the records of main: pico_w_alpha +5,184 B used and +4,096 B .bin, pico_wire +4,680 B used and .bin, RAM 125,788 B and 104,592 B. The budget goes down with the image, and the OTA table (PLANO_STABLE 3.1) says the same. chore/v2.14.0, 2026-11-05: pico_w_release 1,024,828 -> 1,024,820 B used (-8), .bin 1,036,860 -> 1,036,852 (-8); pico_w_test 1,025,644 -> 1,025,636 B used (-8), .bin 1,037,676 -> 1,037,668 (-8); pico_w_asserts 1,027,124 -> 1,027,116 B used (-8), .bin 1,039,156 -> 1,039,148 (-8); pico_wire 597,212 -> 597,204 B used (-8), .bin 609,248 -> 609,240 (-8); pico_wiretft 776,164 -> 776,156 B used (-8), .bin 788,200 -> 788,192 (-8); pico_wiretft_angulo 817,140 -> 817,132 B used (-8), .bin 829,176 -> 829,168 (-8) (SIMUT_VERSION 2.13.0 -> 2.14.0, the only source change; the login page's gzip member packs 6,666 -> 6,660 B (LOGIN_PAGE_GZ, the only symbol whose size differs in the ELF of any of the eleven) and what follows it sits 8 B closer, so these six images and pico_w_test_https (no .bin record: 1,029,980 -> 1,029,972 B used, -8) lose 8 B, while pico_w_air, pico_w_alpha, pico2_w_release and pico2_w_original keep the same used and the same .bin, their padding absorbing the shift; all eleven built from zero through the pio shim under the shared build lock, -Werror holding, the unchanged tree built the same way first and sha256-equal to main's eleven (11 of 11)).; fix/web-texts-manual-review, 2026-11-06 (565f725c; all eleven from zero through the pio shim under the shared build lock, -Werror holding, no warning from src/; the base 95a1d2ae built the same way, the eleven .bin sizes read from the sibling session's wt_tags_base builds of that commit, equal the records to the byte; used against the base pico_w_release +512, pico_w_test +520, pico_w_test_https +424, pico_w_asserts +512, pico_w_alpha +192, pico_w_air +192, pico2_w_release +520, pico_wire +568, pico_wiretft +520, pico_wiretft_angulo +520, pico2_w_original +520, the .bin of pico_w_alpha and pico_w_air equal, their alignment absorbing the 192 B; the growth is the four fixes together (lang.js, four pages, 27 keys' markup, ScanResultJson.h), not split by fix; every budget and each changed image's .bin recorded); fix/web-texts-manual-review, the review's two commits, 2026-11-06 (32bb8a26: 38681b83 is test code only; 32bb8a26 gives the export box's failure counter a key of its own, exp_failed, in place of the dashboard's dash_metr_err; all eleven from zero again through the pio shim under the shared build lock, -Werror holding, no warning from src/; used against the first build of this branch (565f725c) pico_w_release -8, pico_w_test -8, pico_w_test_https -8, pico_w_asserts -8, pico_wire -8, pico_wiretft -8, pico2_w_original -8, and pico_w_alpha, pico_w_air, pico2_w_release and pico_wiretft_angulo equal; against the base 95a1d2ae the growth is pico_w_release +504, pico_w_test +512, pico_w_test_https +416, pico_w_asserts +504, pico_w_alpha +192, pico_w_air +192, pico2_w_release +520, pico_wire +560, pico_wiretft +512, pico_wiretft_angulo +520, pico2_w_original +512; the seven budgets and .bin that moved recorded, the budgets lowered with the images); fix/web-texts-manual-review, the history page's two fixes, 2026-11-06 (246a2a7c: the Origem selector, a count element of its own and a key on the label only while it shows a word; eacef7be: the Max and Min badges, hist_max and hist_min in both packs, which are files on the device and not part of the image; all eleven from zero again in one pio run through the pio shim under the shared build lock, -Werror holding, no warning from src/; used against the review's build (50a78f05) pico_w_release +160, pico_w_test +160, pico_w_test_https +160, pico_w_asserts +160, pico_wire +160, pico_wiretft +160, pico_wiretft_angulo +152, pico2_w_original +136, pico2_w_release +128, and pico_w_alpha and pico_w_air equal: their .rodata is a multiple of 4,096 B (438,272 and 454,656), so the padding absorbs the page's growth, and the new page is in both ELFs (the history page's gzip member read back out of each, with the new markup in it); against the base 95a1d2ae the growth is pico_w_release +664, pico_w_test +672, pico_w_test_https +576, pico_w_asserts +664, pico_w_alpha +192, pico_w_air +192, pico2_w_release +648, pico_wire +720, pico_wiretft +672, pico_wiretft_angulo +672, pico2_w_original +648; the nine budgets and .bin that moved recorded); fix/web-texts-manual-review, the second review's commits, 2026-11-06 (778cd368: the Origem label's count guard; c674df49: the overlay and the progress lines of the history page written through _setText( ), with their key and English or none; all eleven from zero again in one pio run through the pio shim under the shared build lock, -Werror holding, no warning from src/; used against the history page's first two fixes (eacef7be, recorded in 1b2b4785) pico_w_release +72, pico_w_test +72, pico_w_test_https +72, pico_w_asserts +72, pico2_w_release +104, pico_wire +72, pico_wiretft +72, pico_wiretft_angulo +80, pico2_w_original +96, and pico_w_alpha and pico_w_air equal (their .rodata is still a multiple of 4,096 B, 438,272 and 454,656, which absorbs the growth; the new page is in both ELFs, read back out of each, with _setText( ) in it); against the base 95a1d2ae the growth is pico_w_release +736, pico_w_test +744, pico_w_test_https +648, pico_w_asserts +736, pico_w_alpha +192, pico_w_air +192, pico2_w_release +752, pico_wire +792, pico_wiretft +744, pico_wiretft_angulo +752, pico2_w_original +744; the nine budgets and .bin that moved recorded); feat/screen-boot-tags, 2026-11-06, on 95a1d2ae (before the merge of main): pico_w_test 1,025,636 -> 1,027,884 B used (+2,248), .bin 1,037,668 -> 1,039,916 (+2,248), RAM 109,212 -> 111,060 B (+1,848); pico_w_test_https 1,029,972 -> 1,032,220 B used (+2,248), RAM 109,228 -> 111,076 B (+1,848) (the boot terminal's recorder, `screen boot` and `screen clkboot`, all under SIMUT_BOOT_TAGS, SIMUT_CLI_FULL && SIMUT_DISPLAY_TFT: .text +2,056, .rodata +192, .bss +1,848, .data unchanged; the other nine images sha256-equal to 95a1d2ae, all eleven built from zero through the pio shim on both sides, base and tip each in a worktree of its own, and each one's .bin equal to its record before this change).; feat/screen-boot-tags after the merge of main (#324), 2026-11-06: pico_w_test built from zero through the pio shim under the shared build lock, -Werror holding, no warning from src/: 1,028,628 B used (main 1,026,380, +2,248), .bin 1,040,660 B (main 1,038,412), RAM 111,060 B; the budget recorded as measured and the 'bin' dropped with the exemption; the nine images without the tags, built the same way, equal main's records to the byte in used and in .bin.; feat/screen-boot-tags with the tags in pico_w_test only (019a4996), 2026-11-06: all eleven from zero in one pio run through the pio shim under the shared build lock, -Werror holding, no warning from src/, and a clean copy of main (3b09e9ff) built the same way beside them; the ten images without the tags are sha256-equal to main's, pico_w_test_https included (1,030,620 B used, 1,042,652 B .bin, as main; with the tags it did not link, 420 B over the program slot), and pico_w_test is +2,248 B used and +2,248 B .bin against main (1,028,628 and 1,040,660 B); no budget and no 'bin' moves.; chore/repo-hygiene-2026-11-07 on 6c8fff60 plus this change, 2026-11-07 (all eleven built from zero through the pio shim of 2026-10-28 against ~/.platformio, no other build running, -Werror holding, no warning from src/: the login page's tagline gains 'Integrado' and the Angulo's boot subtitle changes; used against 6c8fff60 +8 on pico_w_release, pico_w_test, pico_w_test_https, pico_w_asserts, pico_wire, pico_wiretft and pico2_w_original, the .bin of each +8; pico_w_alpha, pico_w_air, pico2_w_release and pico_wiretft_angulo equal in used and in .bin; the seven budgets and five .bin recorded; built again from zero after the review's comment-only edits to 39 src/ headers, WebUI.h and platformio.ini: 11 of 11 equal in used and in .bin, the new subtitle in the two Angulo .bin and the tagline with 'Integrado' in the login page of pico2_w_release as in pico_w_release's)",
"measured_at_commit": "2026-10-13 feat/wire-driver 4912762 (the seven radio images; the budgets only move here), feat/wire-surface, the log-code commit, and feat/net-boundary, B2 on 20c06b0 plus this change, then ebcf29e plus the review's fixes (pico_wire, pico_wiretft); feat/net-radio-unit on 3055b31 plus this change (all nine); fix/radio-link-in-ntp-wait on 53b6d7b plus this change (the seven radio images).; fix/silent-radio-no-blind-join on 2290638 plus this change (all nine).; fix/link-health-in-ntp-wait on 5293bae plus this change (all nine). measured is the zopfli image PlatformIO prints.; fix/web-short-write on f1b8b8f plus this change (all nine), built against a PlatformIO core that carries the override (PLATFORMIO_CORE_DIR, a hardlinked copy of ~/.platformio patched by patch.sh).; fix/web-fewer-writes on 686a1de plus this change (all nine), built against a PlatformIO core that carries overrides 2m and 2n (PLATFORMIO_CORE_DIR, a hardlinked copy of ~/.platformio patched by patch.sh).; deps/pico-enc28j60-1.3.1 on 192e1ad plus this change (all nine), built against ~/.platformio, which carries overrides 2m and 2n since 2026-10-15.; fix/wire-tft-flash-render-274 on d5f7063 plus this change (the five HTTPS images), built against ~/.platformio patched by patch.sh, which carries overrides 2m, 2n and the WiFiServerSecure accept patch (2b-ter).; perf/tft-spi-clk-peri on 80c3c88 plus this change (the six panel images), built against the ~/.platformio as main left it (its pico_w_release .bin matched the record to the byte before the change).; feat/ui-drivers S2b-3 on 0c6a742 plus this change (pico_wiretft_angulo, joining; pico_wiretft and pico_w_release rebuilt as controls and equal to their records), built against a PlatformIO core that carries overrides 2m and 2n (PLATFORMIO_CORE_DIR, a private copy of ~/.platformio patched by patch.sh).; feat/ui-drivers-s5 (S5) on 60eabd6 and again on 33836c2 (the nine product images and pico_wiretft_angulo, equal at both), built against a PlatformIO core that carries overrides 2m and 2n (PLATFORMIO_CORE_DIR, a private copy of ~/.platformio patched by patch.sh).; feat/ui-drivers-s5 review (C10c) on a1fda61 plus this change (pico_wiretft_angulo), built against a PlatformIO core that carries overrides 2m and 2n (PLATFORMIO_CORE_DIR, a private copy of ~/.platformio patched by patch.sh).; feat/ui-angulo-visual on 2c89d12 and on this change (pico_wiretft_angulo), built against a PlatformIO core that carries overrides 2m and 2n (PLATFORMIO_CORE_DIR, a private copy of ~/.platformio patched by patch.sh).; feat/ui-drivers-s6 (S6) on 7ee79e9 (811,064 B), a872d93 (811,192 B, .bin 823,224), 77e89aa (811,168 B, .bin 823,200), 390fc6f, 05da4f0, 80ecdb0 (811,232 B, .bin 823,264), 5da8888 and c078ef4 (811,232 B, .bin 823,264): pico_wiretft_angulo, and the nine product images equal to their records at every step, built against a PlatformIO core that carries overrides 2m and 2n (PLATFORMIO_CORE_DIR, a private copy of ~/.platformio patched by patch.sh); feat/ui-drivers-s7 (S7): the Angulo image measured at C3 43960a7 (+80) and C4 64dc6c6 (-8) plus each commit's own change, and unchanged at C5 af96f64, C6 6102e9f, C7 5e27b04; the nine product images keep their records.fix/ui-angulo-aviso on e421a9c plus this change (pico_wiretft_angulo only: 811,656 B, .bin 823,688; PLATFORMIO_CORE_DIR as above, no patch.sh run). Then the web notice as \"Web: \" on 3abff93 plus this change (pico_wiretft_angulo only: 811,640 B, .bin 823,672). fix/ui-angulo-toque on a6026d5 plus this change (pico_wiretft_angulo only: 811,928 B, .bin 823,960; PLATFORMIO_CORE_DIR as above, no patch.sh run). Then fix/ui-angulo-historico merged over it (pico_wiretft_angulo only: 811,952 B, .bin 823,984; PLATFORMIO_CORE_DIR as above, no patch.sh run). fix/ui-angulo-navegacao on b035e92 (pico_wiretft_angulo only: 812,352 B, .bin 824,384; PLATFORMIO_CORE_DIR as above, no patch.sh run). fix/ui-angulo-navegacao review on 9d3e8c9 (pico_wiretft_angulo only: 812,544 B, .bin 824,576; PLATFORMIO_CORE_DIR as above, no patch.sh run).Then fix/ui-angulo-historico, second version, on 4122282 plus this change (pico_wiretft_angulo only: 811,792 B, .bin 823,824; PLATFORMIO_CORE_DIR as above, no patch.sh run). Then fix/ui-angulo-historico 456a1ad merged over 91df450 (pico_wiretft_angulo only: 812,600 B, .bin 824,632; PLATFORMIO_CORE_DIR as above, no patch.sh run). S8a, feat/ui-drivers-s8a at C8 (the ten images; PLATFORMIO_CORE_DIR as above, no patch.sh run): stages/s8a_c8. fix/ui-angulo-margem on 3ba3ea5 plus this change (pico_wiretft_angulo only: 813,152 B, .bin 825,184; PLATFORMIO_CORE_DIR as above, no patch.sh run): stages/margem. perf/ui-angulo-pintura on 89de38d plus this change (pico_wiretft_angulo, p1). perf/ui-angulo-pintura on ed4af6f plus this change (pico_wiretft_angulo, p2). perf/ui-angulo-pintura on 24646e0 plus this change (pico_wiretft_angulo, p3). perf/ui-angulo-pintura on 536f5e7 plus this change (pico_wiretft_angulo, p4). perf/ui-angulo-pintura on 4cc3e71 plus this change (pico_wiretft_angulo, p5). perf/ui-angulo-pintura on b3724ce plus this change (pico_wiretft_angulo, p6). perf/ui-angulo-pintura on fbc5d73 plus this change (pico_wiretft_angulo, p7). perf/ui-angulo-pintura on 902db57 plus this change (pico_wiretft_angulo, p8). perf/ui-angulo-pintura on a5d3333 plus this change (pico_wiretft_angulo, p9). perf/ui-angulo-pintura on 233b9c3 plus this change (pico_wiretft_angulo, p10). perf/ui-angulo-pintura on 4ee3b4b plus this change (pico_wiretft_angulo, p11a). perf/ui-angulo-pintura on d71a313 plus this change (pico_wiretft_angulo, p11b). perf/ui-angulo-pintura on 56a7a35 plus this change (pico_wiretft_angulo, p12a).; sync/ui-drivers-main (the merge of main 2d6650b into feat/ui-drivers 1a913a4) plus this change (the nine product images and pico_wiretft_angulo), built against ~/.platformio, which carries overrides 2m, 2n and 2b-ter and the rest of patch.sh (no PLATFORMIO_CORE_DIR).; sync2/ui-drivers-main on the merge of 90e82d1 into d9cdda4 (all ten images, from scratch, default core).; fix/safe-reboot-writes-pending-log on 90e82d1 plus this change (all nine, from scratch), built against ~/.platformio, which carries every override main's patch.sh applies.; fix/radio-answers-by-return-code rebased on 7e3f29a6 plus this change (all nine, from scratch), built against ~/.platformio, which carries every override main's patch.sh applies.; sync3/ui-drivers-main, the merge of feat/ui-drivers 154ff60c and main 2f4bc3d3 (the ten images built from scratch on the merge tree, whose five measured files held the branch's text meanwhile and do not reach the build), built against ~/.platformio, the default core.; s10a/c2, the limits editor's step in form Fb (the ten images built from scratch on the committed tree by tools_s10a/verify_step.sh; the three LCD images are byte for byte C0's), built against ~/.platformio, the default core.; fix/language-pack-log-records on 2f4bc3d3 plus this change (all nine, from scratch), built against ~/.platformio, which carries every override main's patch.sh applies.; sync4/ui-drivers-main, the merge of feat/ui-drivers edc95d6d and main 25e169df (the ten images built from scratch on the merge tree, whose six conflicting files held the branch's text meanwhile and do not reach the build), built against ~/.platformio, the default core.; s10b/c0b on 50828bdf plus this change (the ten images, then pico2_w_angulo, all from scratch with tools/build9.sh), built against ~/.platformio, the default core.; fix/alpha-console-event-names on 25e169df plus this change (all nine, from scratch), built against ~/.platformio, which carries every override main's patch.sh applies.; chore/https-test-image-room on a8ccf4c3 plus this change (all nine, from scratch), built against ~/.platformio, which carries every override main's patch.sh applies.; sync5/ui-drivers-main, the merge of feat/ui-drivers ecb527e3 and main 2075b84e (the eleven images built from scratch on the merge tree, whose five conflicting files were resolved before the build began; the four that are not the language-pack gate do not reach the image), built against ~/.platformio, the default core.; fix/lang-english-way-back on be56633d plus this change (all nine, from scratch), built against ~/.platformio, which carries every override main's patch.sh applies.; fix/panel-theme-lang-permission on d2c366b3 plus this change (all nine, from scratch, before and after), built against ~/.platformio with PlatformIO 6.1.19 and zopfli 0.4.3 in a fresh Python 3.11 venv, because the pipx venv stopped working when /usr/bin/python3 went from 3.12 to 3.14 (2026-10-24, dpkg.log); on d2c366b3, before the change, that venv gave the eight recorded .bin sizes to the byte, and a Python 3.14 venv over the same PlatformIO sources gave the same sha256 on all nine images of both trees.; sync6/ui-drivers-main, the merge of feat/ui-drivers 5f2cdafd and main 9f4da77b (tree 3e55b57b; the eleven images built from scratch on the merge tree, whose twelve conflicting files were resolved before the build began, and once more on the staged tree the commit records, 11 of 11 sha256-equal; the two Angulo images rebuilt from scratch after their Idioma screen took #294, the nine products sha256-equal to the merge's, and the eleven again after the docs fixes, 11 of 11 equal; the file's numbers are these builds': the nine from the merge tree, the two Angulo rows from the rebuild), built against ~/.platformio, the default core, with PlatformIO 6.1.19 through the shim of stages/s10c_d6_0: a Python 3.14 venv over the pipx venv's PlatformIO sources, with the apt python3-zopfli 0.4.1 where the pipx one had 0.4.3, because the pipx venv stopped working when /usr/bin/python3 went from 3.12 to 3.14 (2026-10-24, dpkg.log); its pico_wire gave the base's sha256 (57721645) to the byte before any image was trusted.; chore/test-image-license-fs on 9f4da77b plus this change (all nine, from scratch), built through the pio shim of 2026-10-28 (the pipx pio is dead since the python upgrade) against ~/.platformio; the eight others come out byte for byte main's.; sync6b/ui-drivers-main, the merge of feat/ui-drivers abf43e33 and main 99e9001b (#297, pico_w_test serves its licence page from LittleFS; the eleven images built from scratch on the merge tree, whose two conflicts, PLANO_STABLE.md and this file, were resolved before the build began, while tools/generated/profiles.ini merged clean and gen_features.py --check agrees with it; ten of them sha256-equal to abf43e33's, pico_w_test the only one that changed, by the three symbols main's own pair moves), built against ~/.platformio, the default core, through the same pio shim as the sync6 entry above.; feat/web-theme-names on 99e9001b plus this change (all nine, from scratch, built through the pio shim of 2026-10-28 against ~/.platformio; origin/main built the same way gave every recorded .bin size to the byte).; feat/ui-drivers sync 7 (the merge of main c9028947, #298, into 290c2662; all eleven from scratch through the pio shim of 2026-10-28, the baseline 290c2662 built the same way first: ten images moved as main's pair did for their family, pico_w_alpha and pico_w_air by sha256 only; pico_w_test_https 1,029,020 -> 1,029,188 B used (+168); each image's .bin recorded); feat/ui-drivers sync 8 (the merge of main cf97b955, #299..#302, into 22d83de3; all eleven from scratch through the pio shim of 2026-10-28, the baseline 22d83de3 built the same way first and sha256-equal to sync 7's in all eleven: the eight page images moved by FILE_PAGE_GZ and LOGIN_PAGE_GZ alone, as main's pair did for their family, pico_w_alpha and pico_w_air by sha256 only, pico_w_test_https not at all; each image's .bin recorded) 2026-10-31 fix/alarm-limits-snapshot on 3ca4976f: the eight TFT images' budgets and .bin, all eleven rebuilt from zero here.; fix/display-offset-leave, 2026-10-31 (aaccce35, the F3 fix, merged with the display-alignment fix; all eleven from zero through the pio shim of 2026-10-28 under the shared build lock, -Werror holding; used against aaccce35 pico_w_release +216, pico_w_test +216, pico_w_test_https +216, pico_w_asserts +224, pico_w_alpha +8, pico_w_air +8, pico2_w_release +200, pico_wire +8, pico_wiretft +224, pico_wiretft_angulo +224, pico2_w_original +200; each changed image's .bin recorded); fix/keypad-route-angulo, 2026-10-31 (6981fd93 plus this change; all eleven from zero through the pio shim of 2026-10-28 under the shared build lock, -Werror holding; the baseline 6981fd93 built the same way first and equal to every recorded budget and .bin; the nine images that are not the Angulo's have the same sha256 as the baseline's; used against 6981fd93 pico2_w_release -208 and pico_wiretft_angulo -216; the two .bin recorded); fix/small-bugs, 2026-11-04 (2128deb1; all eleven from zero through the pio shim under the shared build lock, -Werror holding; the base 92103fb5 built the same way with the linker map, and each fix's share measured as an A/B on the images it changes; each changed image's .bin recorded); fix/small-bugs, 2026-11-04 (212ff4bd, recorded again after the review; all eleven from zero through the pio shim under the shared build lock, -Werror holding); feat/alpha-lcd-print, 2026-11-04 (34fc44a9 plus this change: loopCore1( ) of the alpha display draws through display/AlphaFrame.h; all eleven from zero through the pio shim of 2026-10-28 under the shared build lock, -Werror holding; the baseline 34fc44a9 built the same way first, and pico_w_alpha and pico_wire built from zero twice there with the same sha256; the nine images that do not compile DisplayManager_Alpha.cpp have the same sha256 as the baseline's; used against 34fc44a9 pico_w_alpha +32 and pico_wire +24, the .bin of pico_wire +24 and the .bin of pico_w_alpha equal, its alignment absorbing the 32 B; both budgets and the pico_wire .bin recorded); feat/alpha-lcd-print after the merge of main (#318) at b0171d1e, 2026-11-04: all eleven from zero through the pio shim under the shared build lock, the nine images without the alpha's LCD sha256-equal to main's (d8871388, the eleven of #318's last build), used against main pico_w_alpha +24 and pico_wire +32, the .bin of pico_wire +32 and the .bin of pico_w_alpha equal; both budgets and the pico_wire .bin recorded; feat/screen-boot-tags on 5b191cbe plus this change, 2026-11-06 (all eleven), built from zero through the pio shim against ~/.platformio as it stands, the base 95a1d2ae built the same way in a worktree of its own: the nine images without the tags sha256-equal to it, pico_w_test and pico_w_test_https +2,248 B used, both budgets and pico_w_test's .bin recorded; feat/screen-boot-tags on 6e70a10f (the merge of main at 3b09e9ff) plus this change, 2026-11-06 (ten of the eleven built from zero through the pio shim under the shared build lock: pico_w_test_https does not link there, 420 B over the program slot), pico_w_test's budget recorded and its 'bin' dropped with the exemption; feat/screen-boot-tags on 019a4996 plus this change, 2026-11-06 (all eleven built from zero through the pio shim under the shared build lock, beside a clean build of main at 3b09e9ff: the ten images without the tags sha256-equal to it, pico_w_test +2,248 B used, nothing recorded moves); chore/repo-hygiene-2026-11-07 on 6c8fff60 plus this change (all eleven, from zero), built through the pio shim of 2026-10-28 against ~/.platformio.",
"ota_exempt": {
"pico_w_test": "A bench image: the one with the full console and the TFT panel, which the rig runs for the suites (AGENTS.md). It reaches the rig over USB, and release-ota.yml does not publish it. On 2026-11-06 the boot tags (#325, `screen boot` and `screen clkboot`: +2,248 B of .bin) and the web texts (#324: +744 B) together took its signed .bin to 1,040,901 B, 517 B over OTA_APP_SAFE_MAX_SIZE (1,040,384 B); it was 227 B under with the tags alone, and 1,731 B under on main with the texts alone. Over the ceiling the device refuses to stage it, which costs only installing this image over the air; its flash budget still holds. Exempt since 2026-11-06, the maintainer's decision, as pico_w_test_https has been since 2026-10-02 (docs/analysis/PLANO_STABLE.md, B6).",