You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The 2 failing runs in the monitored window were both PR runs whose ci.yml carried a Workflow regression guards (workflow-tests) job (introduced by in-flight PR #429). Its Install yq step died deterministically with:
sha256sum: 'standard input': no properly formatted checksum lines found
The mikefarah/yq checksums release asset is a multi-hash table: the asset filename is column 1 (at the start of each line) and SHA-256 is column 19 (per checksums_hashes_order). The buggy step used grep "yq_linux_amd64$" | awk '{print $1}' β
an end-anchored match that never matches (filename is at line start), yielding an empty
checksum and a hard failure. (Verified: column 19 == the real SHA-256 of yq_linux_amd64.)
Scope
Lock the correct yq-checksum-extraction invariant into ci.yml's existing Fleet Monitor
regression guard (scripts/test-ci-workflow.sh, from #380) so this exact drift can never
recur, following the established per-workflow guard convention (#374, #380, #428). No change
to ci.yml itself is made here (the buggy step is not on main; PR #429 already carries the
corrected $19 extraction) β the guard enforces correctness whenever that job lands.
scripts/test-ci-workflow.sh β add Check 5: if an Install yq step is present, its checksums extraction must anchor the filename at line start and select the SHA-256
column ($19), never the filename column ($1); skips cleanly when no yq step exists
Both failing runs in the monitored window were PR runs whose ci.yml carried a Workflow regression guards (workflow-tests) job introduced by in-flight PR #429. Its Install yq step failed deterministically with sha256sum: 'standard input': no properly formatted checksum lines found. The mikefarah/yq checksums asset is a multi-hash table β
the asset filename is column 1 (at line start) and SHA-256 is column 19 (per checksums_hashes_order, verified against the real binary hash). The buggy step used grep "yq_linux_amd64$" | awk '{print $1}': the end-anchor never matches and column 1 is the
filename, so EXPECTED was empty. secret-scan (the only job on main) was green throughout.
Plan Execution
Diagnosed the two failing runs β Install yq checksum-extraction bug
Verified column 19 == SHA-256 of yq_linux_amd64 from the live release asset
Wrote fixture-based tests first (red), then implemented (green)
Added optional-arg support + Check 5 to the existing ci.yml regression guard
Ran required lint (dev-lead-lint.sh), shellcheck, and every repo guard/test script
Test Results
ok - buggy yq extraction is rejected
ok - correct yq extraction (^yq_linux_amd64 / column 19) is accepted
ok - workflow without an Install yq step passes (check skipped)
ok - workflow missing a concurrency block is rejected
All tests passed.
$ bash scripts/test-ci-workflow.sh # against real ci.yml
PASS: no 'Install yq' step in .github/workflows/ci.yml β yq checksum check not applicable
All checks passed.
$ bash .dev-lead/scripts/dev-lead-lint.sh
[lint] validating CODEOWNERS...
OK: CODEOWNERS
[lint] all checks passed
$ shellcheck --severity=warning -x scripts/test-ci-workflow*.sh
clean
Files Changed
scripts/test-ci-workflow.sh: accept an optional workflow-path arg (mirrors sibling test-dependabot-rebase-workflow.sh); add Check 5 β if an Install yq step is present,
its checksums extraction must anchor the filename at line start and select the SHA-256
column ($19), never an end-anchor or column 1. No-op when no yq step exists.
scripts/test-ci-workflow.test.sh (new): fixture-based red/green tests covering the buggy
extraction (rejected), the correct extraction (accepted), the no-yq skip path, and the
existing concurrency invariant.
The gitleaks step's own awk '{print $1}' is intentionally untouched and not flagged β
Check 5 is scoped to the Install yq block, and gitleaks' checksums.txt genuinely lists
the hash in column 1.
No commits/PRs opened, per the dev-lead protocol β CI handles git operations.
Activity
don-petry commentedon Aug 13, 2026
Dev-Lead Implementation Plan
Issue: #430 β [Fleet Monitor] petry-projects/TalkTerm β .github/workflows/ci.yml
Root Cause
The 2 failing runs in the monitored window were both PR runs whose
ci.ymlcarried aWorkflow regression guards(workflow-tests) job (introduced by in-flight PR #429). ItsInstall yqstep died deterministically with:The mikefarah/yq
checksumsrelease asset is a multi-hash table: the asset filename iscolumn 1 (at the start of each line) and SHA-256 is column 19 (per
checksums_hashes_order). The buggy step usedgrep "yq_linux_amd64$" | awk '{print $1}'βan end-anchored match that never matches (filename is at line start), yielding an empty
checksum and a hard failure. (Verified: column 19 == the real SHA-256 of
yq_linux_amd64.)Scope
Lock the correct yq-checksum-extraction invariant into
ci.yml's existing Fleet Monitorregression guard (
scripts/test-ci-workflow.sh, from #380) so this exact drift can neverrecur, following the established per-workflow guard convention (#374, #380, #428). No change
to
ci.ymlitself is made here (the buggy step is not onmain; PR #429 already carries thecorrected
$19extraction) β the guard enforces correctness whenever that job lands.Implementation Checklist
scripts/test-ci-workflow.shβ accept optionalWORKFLOWarg (mirrors siblingtest-dependabot-rebase-workflow.sh) to enable fixture testingscripts/test-ci-workflow.shβ add Check 5: if anInstall yqstep is present, itschecksumsextraction must anchor the filename at line start and select the SHA-256column (
$19), never the filename column ($1); skips cleanly when no yq step existsTests to Write
scripts/test-ci-workflow.test.shβ fixture-based red/green tests:yq_linux_amd64$/$1) β guard FAILS (red)^yq_linux_amd64/$19) β guard PASSES (green)Test Command
bash scripts/test-ci-workflow.test.sh && bash scripts/test-ci-workflow.shdon-petry commentedon Aug 13, 2026
Dev-Lead: Implementation Complete
Root Cause (confirmed)
Both failing runs in the monitored window were PR runs whose
ci.ymlcarried aWorkflow regression guards(workflow-tests) job introduced by in-flight PR #429. ItsInstall yqstep failed deterministically withsha256sum: 'standard input': no properly formatted checksum lines found. The mikefarah/yqchecksumsasset is a multi-hash table βthe asset filename is column 1 (at line start) and SHA-256 is column 19 (per
checksums_hashes_order, verified against the real binary hash). The buggy step usedgrep "yq_linux_amd64$" | awk '{print $1}': the end-anchor never matches and column 1 is thefilename, so
EXPECTEDwas empty.secret-scan(the only job onmain) was green throughout.Plan Execution
Install yqchecksum-extraction bugyq_linux_amd64from the live release assetdev-lead-lint.sh), shellcheck, and every repo guard/test scriptTest Results
Files Changed
scripts/test-ci-workflow.sh: accept an optional workflow-path arg (mirrors siblingtest-dependabot-rebase-workflow.sh); add Check 5 β if anInstall yqstep is present,its
checksumsextraction must anchor the filename at line start and select the SHA-256column (
$19), never an end-anchor or column 1. No-op when no yq step exists.scripts/test-ci-workflow.test.sh(new): fixture-based red/green tests covering the buggyextraction (rejected), the correct extraction (accepted), the no-yq skip path, and the
existing concurrency invariant.
Notes
ci.ymlitself is not modified: the buggy step is not onmain, and PR feat: implement issue #428 β [Fleet Monitor] petry-projects/TalkTerm β .github/workflows/dependabot-rebase.ymlΒ #429 alreadycarries the corrected
awk '{print $19}'extraction. This guard locks that correctness inso the drift cannot recur whenever the
workflow-testsjob lands β matching the establishedper-workflow Fleet Monitor guard convention ([Fleet Monitor] petry-projects/TalkTerm β .github/workflows/pr-review.ymlΒ #374, [Fleet Monitor] petry-projects/TalkTerm β .github/workflows/ci.ymlΒ #380, [Fleet Monitor] petry-projects/TalkTerm β .github/workflows/dependabot-rebase.ymlΒ #428). The guard was validated
against PR feat: implement issue #428 β [Fleet Monitor] petry-projects/TalkTerm β .github/workflows/dependabot-rebase.ymlΒ #429's exact corrected pattern (accepted).
awk '{print $1}'is intentionally untouched and not flagged βCheck 5 is scoped to the
Install yqblock, and gitleaks'checksums.txtgenuinely liststhe hash in column 1.
feat: implement issue #430 β [Fleet Monitor] petry-projects/TalkTerm β¦
feat: implement issue #430 β [Fleet Monitor] petry-projects/TalkTerm β¦
feat: implement issue #430 β [Fleet Monitor] petry-projects/TalkTerm β¦
feat: implement issue #430 β [Fleet Monitor] petry-projects/TalkTerm β¦