Dev - #67
Merged
Merged
Conversation
…tion parseNegotiateProtocolVersion() read the server's count of unrecognized startup options but then always read exactly one option-name string regardless of that count - a count of 0 (version mismatch only, no unrecognized options) or 2+ would desync the buffer offset and corrupt every message parsed after it. It now reads exactly `count` strings into NegotiateProtocolVersionMessage.unrecognizedOptions. The message was also parsed but never dispatched anywhere, so a real negotiation reply from the server was silently discarded. PgSocket now stores it and emits a 'debug' event; IntlConnection and Connection each expose it as a new readonly protocolNegotiation getter, mirroring the existing processID/secretKey/sessionParameters delegation chain - undefined means the server recognized everything this client's startup packet asked for. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Adds Connection.callFunction(functionId, args, options?), calling a function by OID directly instead of through SQL - superseded in current PostgreSQL by SELECT func(...) over Simple/Extended Query, which is why no other current client bothers with it, but it was one of the last gaps in this driver's own wire-protocol coverage. Arguments and the result are raw wire-format bytes rather than typed JS values, matching the low-level nature of the protocol itself. Also fixes a real parsing bug found while wiring this up: parseFunctionCallResponse read `len - 4` bytes as the result directly, but the result value has its own length prefix ahead of it (-1 for SQL NULL, no bytes following in that case) - the old code folded that prefix into the returned buffer as leading garbage and never supported NULL at all. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Adds DatabaseConnectionParams.longCancelKey (default false): requests protocol 3.2 instead of 3.0, which lets the server hand out a cancellation key up to 256 bytes instead of always exactly 4 - the short key is brute-forceable by an attacker who wants to cancel another session's query. Off by default, and never a connection-breaking choice when on: an older server just replies with NegotiateProtocolVersion naming the highest minor version it supports, and the session proceeds there exactly as if longCancelKey had never been set. BackendKeyData's secretKey is now a Buffer instead of a fixed-width number - it has no length prefix of its own, so parseBackendKeyData reads whatever bytes remain in the message (4 of them before 3.2, up to 256 with it). getCancelRequestMessage relays that Buffer back unchanged rather than assuming 4 bytes. This is a breaking change to the public secretKey getter on PgSocket/IntlConnection/Connection. Verified against the real PostgreSQL 18 test server: longCancelKey:true hands back a key longer than 4 bytes with no protocol negotiation needed, and cancel() still genuinely interrupts a running query with it. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Adds details about new features like long cancellation keys, legacy function call protocol, and graceful protocol renegotiation. Updates version numbers in the feature table for PostgreJS 3.1.0.
… in CI apt-get update fails outright if any configured repo's Release/Packages files disagree, even one we never use - the runner image's preinstalled Google Chrome apt source hit exactly that (a hash sum mismatch on Google's end) and took the postgresql.org repo update down with it, failing the job before PostgreSQL could even be installed. Removing the unused repo first isolates the failure to repos we actually depend on. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Protocol 3.2 (and its longer cancel key) only exists on PG 18+; older servers gracefully negotiate back down to 3.0 and hand back the legacy 4-byte key by design, which was failing the assertion on CI's PG 12/16 jobs.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.