Skip to content

chore(gh-aw): temporarily downgrade compiler to v0.87.1 - #1999

Merged
v1v merged 3 commits into
mainfrom
v1v-gh-aw-upgrade-regression-analysis
Sep 3, 2026
Merged

chore(gh-aw): temporarily downgrade compiler to v0.87.1#1999
v1v merged 3 commits into
mainfrom
v1v-gh-aw-upgrade-regression-analysis

Conversation

@v1v

@v1v v1v commented Sep 3, 2026

Copy link
Copy Markdown
Member

What

Downgrades the gh-aw compiler pin (GH_AW_VERSION / GH_AW_BUILD_VERSION / GH_AW_COMPAT_VERSION) from v0.87.10 to v0.87.1 in Makefile, and recompiles all workflow lock files.

Fixes/addresses #1998

Why

Several agentic workflows (e.g. gh-aw-pr-review.lock.yml, consumed by elastic/docs-builder via @v0) are failing their detection job with:

Threat Detection Engine Failure — The analysis engine could not complete. This is a tooling failure, not a security finding.

Upstream github/gh-aw v0.87.2 enabled external threat detection by default (gh-aw#54111), switching the detection job to an external threat-detect (v0.5.1) binary that has an open upstream bug (github/gh-aw-threat-detection#1014): it reports invalid_report_exhausted even after the model successfully calls threat_detection_result. v0.87.1 predates that switch and uses the older inline copilot-prompt detection path instead, sidestepping the bug entirely.

Full rationale, investigation, and tradeoffs are documented in #1998.

Changes

  • Makefile: bump GH_AW_VERSION / GH_AW_BUILD_VERSION / GH_AW_COMPAT_VERSIONv0.87.1
  • Recompiled all 57 .lock.yml workflows + .github/aw/actions-lock.json via make compile

Known tradeoffs (temporary, see #1998 for details)

  • No timeout-minutes: 90 guard on the agent job (falls back to GitHub's 360-min default)
  • Missing GH_AW_PR_HEAD_BASE_* / GH_AW_PR_HEAD_REPO fork-PR-aware safe-outputs plumbing
  • Older container images (gh-aw-firewall 0.28.1, mcpg v0.4.9, github-mcp-server v1.9.0) — misses v0.87.2v0.87.10 safe-output validation/permission/credential hardening fixes
  • OTLP telemetry env wiring removed (not configured in this repo — no practical impact)

Confirmed NOT affected by this downgrade: the mcp-scripts/ready-to-code-review tool (that rename landed upstream in March 2026, long before v0.83.4) — verified present and correctly registered at v0.87.1.

Validation

  • make compile — 57/57 workflows succeeded, 0 errors, same warning set as v0.87.10
  • make lint — passed clean
  • Confirmed v0.87.1 is not in gh-aw's compat.json blocklist (blocked range is v0.82.8v0.85.3)

Follow-up

Track github/gh-aw-threat-detection#1014 upstream and re-upgrade to a fixed gh-aw release once available, reverting this downgrade.

Bypasses the external threat-detect binary (v0.5.1) that ships in
v0.87.2+ and is affected by github/gh-aw-threat-detection#1014
(invalid_report_exhausted despite a successful threat_detection_result
tool call). v0.87.1 predates the external-detector default-on switch
(#54111) and uses the older inline copilot-prompt-based detection
path instead, sidestepping the bug.

This is a temporary measure — revert once gh-aw-threat-detection#1014
is fixed upstream. Known tradeoffs while on v0.87.1:
- No agent job timeout-minutes guard (falls back to GH Actions'
  360-minute default instead of the 90-minute cap added later)
- Missing GH_AW_PR_HEAD_BASE_*/GH_AW_PR_HEAD_REPO plumbing for
  fork-PR-aware safe-outputs handling
- Older security-hardened container images (gh-aw-firewall 0.28.1,
  mcpg v0.4.9, github-mcp-server v1.9.0) and misses v0.87.2-v0.87.10
  safe-output validation/permission-derivation/credential-hardening
  fixes
- OTLP telemetry env wiring removed (not configured in this repo, so
  no practical impact)

Verified: v0.87.1 is not in gh-aw's compat.json blocklist, all 57
workflows recompile with 0 errors (same warning set as v0.87.10), and
the mcp-scripts `ready-to-code-review` tool remains correctly
registered (that fix predates v0.83.4 and is unaffected by this
downgrade).

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
@v1v
v1v requested a review from fr4nc1sc0-r4m0n September 3, 2026 12:55
@v1v v1v self-assigned this Sep 3, 2026
@github-actions github-actions Bot added the medium_boom Medium PR blast radius; likely benefits from human review label Sep 3, 2026

@fr4nc1sc0-r4m0n fr4nc1sc0-r4m0n left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

@v1v

v1v commented Sep 3, 2026

Copy link
Copy Markdown
Member Author

@copilot resolve the merge conflicts in this pull request

then run the make goals to compile and lint and commit the changes

Copilot AI and others added 2 commits September 3, 2026 13:48
…ression-analysis

# Conflicts:
#	.github/workflows/gh-aw-dependency-review.lock.yml
#	.github/workflows/gh-aw-issue-triage.lock.yml
#	.github/workflows/gh-aw-pr-labeler.lock.yml

Co-authored-by: v1v <2871786+v1v@users.noreply.github.com>
Co-authored-by: v1v <2871786+v1v@users.noreply.github.com>

Copilot AI commented Sep 3, 2026

Copy link
Copy Markdown
Contributor

@copilot resolve the merge conflicts in this pull request

then run the make goals to compile and lint and commit the changes

Done in 571043d (merged main and resolved conflicts) and f631201 (recompiled conflicted lock files). I ran make compile and make lint and both passed.

@v1v
v1v merged commit e34a418 into main Sep 3, 2026
25 checks passed
@v1v
v1v deleted the v1v-gh-aw-upgrade-regression-analysis branch September 3, 2026 14:14
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

medium_boom Medium PR blast radius; likely benefits from human review

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants