Pre-commit /verify fails on a timeout, and the review collector stops reading reviewers' scope lists as findings (GST-3201) (#4527)

Pre-commit /verify now fails a check that times out and runs its checks
on one time budget; a reviewer's scope-list bullets no longer need an
owner ruling to clear; stack tests fail when git can't list a PR's
tests.

Why: each let a result read clean with nothing proving it, or block on
nothing (#4518, #4519). Goal: gates never pass on missing evidence.

Nothing to do; `gstack-checks quick --timeout` is now one budget the
checks share (540 s default).

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Review-Finding: p1/outside-voice#1 reader=outside-voice
Review-Finding: p1/outside-voice#2 reader=outside-voice
This commit is contained in:
Garry Tanandcapy-ai[bot] authored and GitHub committed 2026-11-19 15:17:30 +00:00
1 parent 3cccf074ad
commit bd449234aa
14 files changed
+197 -26

No files matched your search

+4 -1
View File
@@ -1114,7 +1114,10 @@ ones, and the arm lost blockers. `bin/gstack-review-dispatch` closes both ends:
`blocking=0` still collects as CRITICAL, as a summary candidate the lead lowers with a reason. An automatic relabel
was tried and removed (gstack#4353): beside a denial line such as "No other CRITICAL issues.", it let a real
unanchored FIXABLE blocker drop to FIXABLE while the count still agreed. A reader that undercounts a real blocker
therefore loses nothing without a written reason. `verify` re-parses each output the way its
therefore loses nothing without a written reason. The same holds for a reader's scope list (GST-3201): when the valid line declares
`blocking=0`, a single-line list item with no structure (no ` | ` table cell, Severity or Blocks field, or file:line) and no
severity label other than one naming findings as a kind ("a HIGH finding still refused", "P0 or P1 findings") keeps its own
label and is marked a summary candidate, so the lead lowers it with a reason; "Two P0 findings remain unfixed" has the same shape and still blocks until judged. `verify` re-parses each output the way its
render asked, prints `VERIFY_VERDICT: agree=<a> asked=<n>` and one `VERDICT <reader> AGREES|DISAGREES|MISSING|MALFORMED`
row per checked reply. These are informational: a disagreeing count has already added or marked its rows, so
they never make a ledger dirty. When collect prints `agrees=no` or `declared=malformed`, read the reply, then judge the
+3 -1
View File
@@ -1166,7 +1166,9 @@ to the base, with a name ending in `.test`, `_test`, `.spec` or `_spec` and a
JS or TS extension, bun's own rule, so a helper or fixture under `test/` is not
one) on the integrated prefix that includes it, T<k>, in a scratch worktree, and prints
`STACK_TESTS: #<pr> T<k> <n> new test file(s) pass|FAIL (<files>)` or `no new
test files`. A failure is `QUEUE_TRAIN_STACK_RED` (nothing written, pushed or
test files`. When git cannot list the files a PR adds, the line is
`STACK_TESTS: #<pr> T<k> FAIL: cannot list the test files it adds: <git error>`,
since an unread list is not an empty one. A failure is `QUEUE_TRAIN_STACK_RED` (nothing written, pushed or
dispatched): the tests pass on the PR head alone only because nothing before
it in the train had changed what they read, so fix them against that prefix or
reorder with `--order`. The default command is `bun --no-env-file
+1 -1
View File
@@ -670,7 +670,7 @@ The skill no longer installs, and `./setup` removes an existing gstack-owned `in
It asks once before running any repo command (`gstack-checks trust`) and asks again when a command changes. A touched file no test covers is `UNMAPPED`, a selection that ran zero tests is `ZERO_RUN`, and neither counts as green. A missing tool or service is reported, not worked around. It stops after five runs or when the same failure comes back twice, and asks for a focused failing test and a root-cause trace before another fix to a stubborn red. When everything passes, `gstack-checks status` prints `READY_FOR_SHIP tree=<tree>` without rerunning anything, and you run `/ship`. It never reviews, bumps versions, pushes or opens a PR.
Since Claude Code 2.1.286, any installed skill named `verify` makes Claude Code tell the model to run it right before each `git commit` (except docs-only and tests-only commits). When the model calls `/verify` that way, without you typing it or asking for checks, it gets a quick check-only pass instead of the loop: one `gstack-checks quick` call that runs the same checks and selected tests, edits nothing, adds no tests, asks nothing and skips gstack's startup steps, then ends with one line, `VERIFY PASS`, `VERIFY FAIL` or `VERIFY skipped: <reason>`. A skip (no `origin` to select tests against, commands you haven't trusted yet, nothing declared) lets the commit go ahead; a `FAIL` makes the model stop and ask you before committing. The arguments `quick` or `pre-commit` ask for the same pass. `gstack-config set verify_before_commit off` turns the pass into an immediate `VERIFY skipped`; installing with `./setup --prefix` (`/gstack-verify`) removes Claude Code's instruction altogether.
Since Claude Code 2.1.286, any installed skill named `verify` makes Claude Code tell the model to run it right before each `git commit` (except docs-only and tests-only commits). When the model calls `/verify` that way, without you typing it or asking for checks, it gets a quick check-only pass instead of the loop: one `gstack-checks quick` call that runs the same checks and selected tests, edits nothing, adds no tests, asks nothing and skips gstack's startup steps, then ends with one line, `VERIFY PASS`, `VERIFY FAIL` or `VERIFY skipped: <reason>`. A check that runs out of time is a `FAIL` (`<check> TIMEOUT`). The checks share one `--timeout` budget, and the test stage starts with what is left. A skip (no `origin` to select tests against, commands you haven't trusted yet, nothing declared) lets the commit go ahead; a `FAIL` makes the model stop and ask you before committing. The arguments `quick` or `pre-commit` ask for the same pass. `gstack-config set verify_before_commit off` turns the pass into an immediate `VERIFY skipped`; installing with `./setup --prefix` (`/gstack-verify`) removes Claude Code's instruction altogether.
---
+3 -1
View File
@@ -7165,7 +7165,9 @@ rule; a helper or fixture under `test/` is not one) on the integrated prefix tha
T<k> (the base plus every slot up to and including that PR), in a scratch
worktree, and they failed. The `STACK_TESTS: #<pr> T<k> <n> new test file(s)
FAIL (<files>): <failing assertion>` line names the files and the runner's
failing lines. Those tests pass on the PR head alone only because nothing
failing lines. A `STACK_TESTS: #<pr> T<k> FAIL: cannot list the test files it
adds: <git error>` line means git could not list the files the PR adds; fix
what the git error names and rerun the dry run. Otherwise, those tests pass on the PR head alone only because nothing
before it in the train had changed what they read (a PR stacked on another
PR's test files or fixtures hit this on 2026-10-26). Nothing was written,
pushed or dispatched; the ordinary dry run (without `--stack-tests`) does not