Skip to content

drvgettable(): fix leaked SQLite statement / persistent lock when SQL_ATTR_MAX_ROWS cap is hit - #22

Open
hwmaier wants to merge 1 commit into
softace:masterfrom
hwmaier:fix-drvgettable-max-rows-lock-leak
Open

drvgettable(): fix leaked SQLite statement / persistent lock when SQL_ATTR_MAX_ROWS cap is hit#22
hwmaier wants to merge 1 commit into
softace:masterfrom
hwmaier:fix-drvgettable-max-rows-lock-leak

Conversation

@hwmaier

@hwmaier hwmaier commented Aug 15, 2026

Copy link
Copy Markdown

If a statement's row-count cap (SQL_ATTR_MAX_ROWS) is reached before the result set is exhausted, the fetch loop broke out without finalizing or resetting the active sqlite3_stmt. The very next outer- loop iteration then overwrote the only reference to it (tres.stmt = NULL), orphaning a still-open statement that keeps holding its SQLite lock for as long as the connection stays open -- blocking every other process from writing to the file, independent of any BEGIN/COMMIT or autocommit state.

Route the max_rows early-exit through the same finalize-or-reset cleanup already used by the loop's normal completion path instead of skipping it, so the statement (and its lock) is released immediately when the cap is hit.

As a result of this, some applications had issues with concurrent access which this PR now fixes.

The motivation for this PR was to fix a compatibility with Altium which is discussed here:
https://stackoverflow.com/questions/45269190/sqlite-odbc-driver-and-database-locking

… is hit

If a statement's row-count cap (SQL_ATTR_MAX_ROWS) is reached before
the result set is exhausted, the fetch loop broke out without
finalizing or resetting the active sqlite3_stmt. The very next outer-
loop iteration then overwrote the only reference to it (tres.stmt =
NULL), orphaning a still-open statement that keeps holding its SQLite
lock for as long as the connection stays open -- blocking every other
process from writing to the file, independent of any BEGIN/COMMIT or
autocommit state.

Route the max_rows early-exit through the same finalize-or-reset
cleanup already used by the loop's normal completion path instead of
skipping it, so the statement (and its lock) is released immediately
when the cap is hit.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant