Repository navigation
Session Handoff [auto-2028]: Bound the Guard's Stdin Descriptor Before int() (#2028) #2030
Description
Next steps
- Fix A Zero-Padded Stdin Descriptor Crashes the Guard's int() Call #2028 in
host-setup/agent-safety/claude/gh-write-guard.py_redirects_stdin: theint(fd)call on a redirect's descriptor prefix is guarded only bytok.isdecimal(), so a digit string over Python's 4300-digit conversion limit raisesValueErrorand the hook exits non-blocking. Strip leading zeros and compare the remainder (or bound the length) before anyint()call, so a zero-padded descriptor still reads as descriptor 0. Done looks like: a self-test row with a zero-padded stdin descriptor over 4300 digits on awhile readloop is denied exactly as the bare0<form is, and the row fails with the fix reverted. - Drive the pull request to develop per
drive-pr, its body carryingCloses on promotion: #2028, then close this lane out.
External blockers
None known.
Internal dependencies
The parked lane auto-1634 (branch feature/auto-1634, blocked) applies the same zero-strip idea to the timeout signal number in this file. Do not build on that branch or merge its work here. Base on develop and keep this change to the descriptor site, so either lane can land first and the other rebases over a small, separate hunk.
State
No branch, worktree, or pull request yet. Work on feature/auto-2028 in a worktree of its own, based on develop.
Parked decisions
None.
What the last round did
Nothing yet. This lane was opened by the unattended-handoff picker.
What not to repeat
Nothing yet.
New learnings
None yet.
Activity
ptr727 commentedon Sep 29, 2026
Parked on decision #2034.
Done. The fix for #2028 is on
feature/auto-2028, pushed at3e285f38, in #2033 to develop withCloses on promotion: #2028. The descriptor-0 test in_redirects_stdinis nowany(unicodedata.decimal(c) for c in fd), which givesint()'s verdict for everyisdecimal()string without converting it. Two self-test rows pin a 4400-zero descriptor (deny on/dev/zero, allow on a file). Reverting the fix fails them. Twolocal-strict-reviewpasses ran and the second is recorded. The first pass found two introduced findings, both fixed, and one pre-existing defect, filed as #2032.Review state. Checks 10/10,
unresolved=0,suppressed=0. Two Copilot rounds on the head: round 1 found nothing, and round 2's one thread was declined with evidence and resolved. CodeRabbit, prompted, found nothing actionable.pr_review.py statusexits 45,COVERAGE_IS_UNSTATED, because theccr-overview-v2format states no coverage line. Round 1's file table names the only changed file.Worktree. Left standing at
/home/pieter/repos/worktrees/ProjectTemplate-auto-2028, clean, onfeature/auto-2028.What remains. Answer #2034. On option 1, re-run
pr_review.py status 2033to confirm the head is still3e285f38with nothing new, then merge #2033 to develop perdrive-prstep 4 and close this lane out. On option 2, re-request Copilot and dispose of whatever it raises. On option 3, the maintainer drives it.ptr727 commentedon Sep 29, 2026
Lane closed out. Decision #2034 answered (merge as is), and #2033 was squash-merged to develop as dd4ca98. #2028 closes on the next promotion. No successor.