* fix(recovery): reclaim proven-stale reply-only ownership with zero queued backlog
Stuck-session recovery kept reply-only ownership forever when the queued
backlog was empty: isActiveRunProgressStale short-circuits to false at
queueDepth 0, so stale active_reply_work was never reclaimed even after
the durable session became killed. Evaluate staleness for reply-only
ownership without the queue gate (the gate stays for run-handle paths),
so proven-stale reply work expires through the existing abort-and-drain
owner path while global-lane and deferred-maintenance exemptions and
live reply work with recent progress are preserved.
Closes#122227
* fix(recovery): keep maintenance phases out of zero-backlog stale reclaim
The zero-backlog reclaim path (requireQueueBacklog: false) applied to
every reply-only operation, including preflight_compacting and
memory_flushing. Those phases are explicitly recognized as compaction
and may honor a configured timeout above the stale threshold, so a valid
long-running maintenance operation could be force-cleared early.
Restore the queue-backlog guard for the maintenance phases so an
unqueued compaction or memory flush is never reclaimed by this path;
ordinary reply-only ownership keeps the zero-backlog expiry. Adds
regressions for both maintenance phases.
* refactor(plugins): retire deactivate hook alias
* refactor(plugin-sdk): prune retired facade exports
* test(logging): isolate logger test controls
* refactor(logging): internalize file transport controls
* test(plugin-sdk): preserve retired facade coverage
* test(auto-reply): remove stale diagnostic imports
* refactor(logging): delete dead config-read guard
shouldSkipMutatingLoggingConfigRead had no production caller even on main;
it survived the dead-export scan only via logger's testApi re-export. The
test-isolation commit removed that mask, exposing the fossil. Delete the
guard, its test-only re-export, its mock entry, and its dedicated test file.
* refactor(plugin-sdk): retire due compatibility subpaths
* test(plugin-sdk): type group policy predicates
* refactor(plugin-sdk): split removed subpath records
* refactor(secrets): remove retired collector barrel
* test(plugin-sdk): tighten wildcard surface pin
* refactor(plugin-sdk): retire matrix facade metadata
* style(plugin-sdk): format facade metadata
* fix(ci): load channel setup contracts from source
Repair the main-owned regression from 99d662473c (Peter Steinberger): the new env-contract test could consume stale ignored dist metadata instead of the checked-in plugin declaration.
* test(plugin-sdk): refresh API baseline after rebase
* fix(logging): parse diagnostics.stability limit/sinceSeq as strict decimal
normalizeDiagnosticStabilityQuery is the only input boundary of the
diagnostics.stability gateway method (no protocol schema), and its
string branch used Number(): "0x10", "1e3", " 5 ", "+42" were
silently coerced instead of rejected, so malformed clients could
filter snapshots by unintended seq values. String params now require
strict decimal digits, matching the boundary's fail-closed contract.
* fix(logging): reject unsafe integers for numeric stability query inputs
Replace Number.isInteger with parseStrictNonNegativeInteger for the numeric path so MAX_SAFE_INTEGER + 1 (and larger) is rejected uniformly across string and number inputs. Add test cases for MAX_SAFE_INTEGER (accepted) and MAX_SAFE_INTEGER + 1 (rejected) for both limit and sinceSeq.
* fix(logging): use normalization-core parser import
---------
Co-authored-by: Altay <altay@hey.com>
* refactor: consolidate coercion ownership
Centralize four canonical coercion helpers, migrate exact core and plugin duplicates through narrow Plugin SDK facades, and enforce declaration and plugin-normalization ownership boundaries.
The sweep adds eight focused SDK exports while deleting more production and tooling code than it adds. User-visible behavior is unchanged except for safer equivalent object and UI parsing at existing boundaries.
* fix: guard integer option ownership
Register resolveIntegerOption with the canonical function owner and extend the declaration-guard fixture so future local duplicates fail validation.
* fix: keep integer helpers on numeric facade
Remove the unshipped duplicate string-coerce exports and route every affected plugin consumer through the existing number-runtime contract.
* fix: point numeric coercion to number runtime
Make boundary and declaration diagnostics recommend the canonical numeric facade, with failing-before coverage for both guidance paths.
* fix(logging): publish diagnostics ZIP through staged atomic overwrite
A named support export wrote the in-memory ZIP buffer straight to the
final path, so a failed or interrupted overwrite truncated the previous
archive, and an existing file kept its pre-existing (potentially
permissive) mode since writeFile only applies mode on creation.
Route publication through writeExternalFileWithinRoot's sibling staging:
the new archive is written to a private sibling, fsynced, and atomically
renamed over the final path. A failed publication leaves the previous
ZIP untouched with no staging residue, and the rename always installs
the staged file's 0o600 mode. The command now returns the actual
published path.
Fixes#122064
* test(logging): prove failed zip replacement stays atomic
* test(sqlite): create private reliability fixture
---------
Co-authored-by: Peter Steinberger <steipete@gmail.com>
* perf(logging): flush file transport asynchronously off the request path
* fix(ci): record flushLogger surface growth and drain before log reads
- +1 public export/callable budget with annotation: logger file-transport
flush for graceful shutdown drains; regenerate SDK API baseline
- three tests that read the log file right after logging now drain the
async file transport first (flushLogger) instead of racing the flusher
* test: drain async log transport before reading talk and request-trace logs
* test: drain async log transport in CLI logger file tests
* test: drop unused withTempDirSync import
* fix(agents): make global loop breaker reachable
Record loop-detector vetoes as typed no-progress outcomes so repeated blocked calls continue the existing streak without colliding with plugin or approval denials. Extract streak accounting into its own owner module and keep completed veto records out of argument reconciliation.\n\nFixes #109435.
* test: harden isolated project routing
Route registry-sensitive UI tests through the isolated project for both focused and broad runs, centralize the isolated file list, and register the Codex prewarm test in the full extension shard.
* test(qa): prove global loop breaker runtime
Drive 31 identical read attempts through the real QA Gateway agent loop and verify the typed veto streak reaches the global circuit breaker before the turn returns a final marker.
* test: keep isolated UI files out of shared runs
Always exclude registry-sensitive files from the shared UI project and reject broad watch targets that would span shared and isolated projects.
Peter Steinberger