og-linuxgsm/.github/pull_request_template.md at master

mirror of https://github.com/GameServerManagers/LinuxGSM synced 2026-08-18 20:23:04 -04:00

chore: unify issue labeling automation and reduce false positives

chore(deps): bump webfactory/ssh-agent from 0.9.0 to 0.10.0 (#4910)

Bumps [webfactory/ssh-agent](https://github.com/webfactory/ssh-agent) from 0.9.0 to 0.10.0.
- [Release notes](https://github.com/webfactory/ssh-agent/releases)
- [Changelog](https://github.com/webfactory/ssh-agent/blob/master/CHANGELOG.md)
- [Commits](https://github.com/webfactory/ssh-agent/compare/v0.9.0...v0.10.0)

---
updated-dependencies:
- dependency-name: webfactory/ssh-agent
dependency-version: 0.10.0
dependency-type: direct:production
update-type: version-update:semver-minor
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
chore(deps): bump dessant/lock-threads from 5 to 6 (#4909)

Bumps [dessant/lock-threads](https://github.com/dessant/lock-threads) from 5 to 6.
- [Release notes](https://github.com/dessant/lock-threads/releases)
- [Changelog](https://github.com/dessant/lock-threads/blob/main/CHANGELOG.md)
- [Commits](https://github.com/dessant/lock-threads/compare/v5...v6)

---
updated-dependencies:
- dependency-name: dessant/lock-threads
dependency-version: '6'
dependency-type: direct:production
update-type: version-update:semver-major
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
feat: sync GitHub issue types with `type:` labels

feat: enable manual backfill for issue labels

This change introduces a `workflow_dispatch` triggered job that allows for manually re-evaluating and applying labels to existing issues.

The new job iterates through a specified set of issues (filtered by state and an optional limit), adds a temporary `maintenance: relabel-backfill` label, and immediately removes it. The primary `labeling-issues` workflow is updated to specifically *not* ignore these bot-generated label events when the temporary label is involved, thus forcing a re-evaluation of the issue by the current labeling logic.

This provides a mechanism to correct mislabeled issues or apply updated labeling rules to the entire issue backlog.

feat: enhance relabel backfill with detailed summary and update github-script

This change significantly improves the observability of the manual issue relabel backfill workflow by introducing:

- Detailed console logging for each issue being processed.
- A comprehensive `core.summary` output in the workflow run, providing tables for overall statistics, processed issues, and any encountered failures.

Additionally, the `actions/github-script` action is updated to v8.

chore: make issue type inference more robust by respecting existing labels

Previously, the automated issue type inference relied solely on the issue title. This could result in valid `type:` labels being removed if the title was ambiguous or didn't explicitly match a predefined pattern.

This change introduces a fallback mechanism where existing `type:` labels are considered if a type cannot be clearly inferred from the title. This reduces unnecessary label churn and improves the accuracy of automated labeling.

fix: backfill runs triage inline instead of label-toggle

GitHub does not fire new workflow runs when GITHUB_TOKEN creates label
events (built-in loop prevention). Replace the label-toggle backfill
approach with an inline version that mirrors the deterministic
reconciliation logic from issue-ai-maintenance directly: type labels,
command/distro labels, game/engine labels, Issue Type GraphQL sync,
and tmux false-positive cleanup. No AI call is made during backfill.

Also removes the now-unnecessary BACKFILL_TRIGGER_LABEL exemption from
the bot-loop guard in issue-ai-maintenance.

fix: track issue type + locked state in backfill summary

- Add 'Issue Type set' column to processed issues summary table
- Track issueTypeSet per issue (null when already correct/unchanged)
- Skip REST label mutations for locked issues (they return 403) with
a console note; Issue Type GraphQL sync still runs for locked issues
- Show lock emoji in issue number column when issue is locked
- Track actual applied add/remove counts (not desired counts)

fix: allow backfill label updates on locked issues

Remove locked-issue skip branch in backfill so label add/remove operations
run for locked issues as well. Keep lock marker in summary for visibility.

fix: keep legacy server request issues classified correctly

Recognize 'server request' anywhere in issue titles (e.g. '[callofduty1] Server Request')
and prefer type: game server request over generic feature when both labels exist.
Apply this in both issue-ai-maintenance and backfill logic.

fix: classify legacy server-request titles in backfill and maintenance

Detect legacy server-request phrasing in issue titles:
- bracketed game/server titles ending with 'Creation'
- titles containing 'Server Creation'
- titles containing 'Server Support' or 'Support for ... server'

Apply the same heuristics in both issue-ai-maintenance and backfill
inferTypeFromTitle paths so old issues are not downgraded to feature.

feat: infer game labels from legacy issue text in relabel

When no structured Game form section is present, infer game labels/scripts
deterministically from title/body using serverlist alias mappings.
Apply this to both issue-ai-maintenance and backfill to improve
historical game labeling without relying on AI.

feat: add optional AI fallback for backfill game detection

Add workflow_dispatch input ai_game_fallback (default false).
In backfill mode, only call AI when deterministic game mapping finds no
match; accept only high-confidence results and map through known game
aliases/labels. Include AI usage stats in the run summary table.

fix: avoid pruning legacy game labels without structured game input

Only remove existing game:* labels when an issue has explicit structured
Game form selections. For legacy title/body inference (and AI fallback),
add matched game labels but do not remove other existing game labels.
This prevents edge cases like issue #1 from losing valid multi-game tags.

fix: require alias evidence for AI game fallback labels

Backfill AI game fallback now accepts a detected game only when the issue
text contains a literal alias token for the mapped game label. Add explicit
logs for AI accept/reject/unmapped outcomes to make attribution auditable
in job logs and prevent false positives like issue #17.

feat: annotate game label adds with detection source in backfill logs

Each game label add now shows its source in the per-issue log line:
#240: added "game: Opposing Force" (text-match)
#248: added "game: Counter-Strike: Global Offensive" (form-field)
#N: added "game: X" (ai-fallback)

Non-game labels (engine, type, needs, etc.) are unchanged.

fix: add missing hasAliasHitForLabel to backfill script context

Each github-script step runs in its own isolated JS context. The backfill
step was calling hasAliasHitForLabel (used by the AI alias-evidence gate)
but the function was only defined in the triage step, causing a ReferenceError
on any issue that triggered AI fallback.

fix: retry AI fallback once on HTTP 429 with Retry-After backoff

When the GitHub Models API rate-limits the backfill (429), read the
Retry-After header (capped at 60s), wait, then retry the request once.
If the retry also fails the issue is skipped as before.

fix: accept joined-token alias evidence in AI game fallback gate

Alias evidence now allows multi-token aliases to match when words are
joined in issue text (e.g. counterstrike vs counter strike), while
keeping exact token checks for single-word aliases.

fix: treat generic AI detections as non-game in backfill

When AI fallback returns generic platform/engine terms (e.g. srcds,
source dedicated server, steamcmd), treat them as non-game detections
instead of logging them as unmapped games. Also prompt the model to
return null for generic terms.

chore: log AI rate-limit headers and 429 count in backfill

Capture Retry-After, X-RateLimit-* and request id on 429 responses,
log them on retry and final skip, and include total AI 429 hits in the
workflow summary table.

fix: disable AI fallback for run on long Retry-After cooldown

When GitHub Models returns 429 with a large Retry-After (over 300s),
stop AI fallback for the remainder of the backfill run instead of
sleeping and retrying per issue. Include disable reason in summary.

fix: prevent overlapping game alias double-matches in text detection

Prefer longest non-overlapping alias matches so titles like
"Killing Floor 2" do not also infer "Killing Floor" unless both
are explicitly present as separate mentions.

fix: prune stale broad game labels when specific game is inferred

For legacy issues without structured game selection, remove existing game
labels only when they are broader overlaps of a newly inferred specific
game label (e.g. remove Killing Floor when Killing Floor 2 is inferred).

fix: stop relabel backfill early when API rate limit is hit

Detect GitHub API rate limit errors during processing, stop the run
gracefully, and report header-derived rate limit details in logs and
summary instead of emitting repeated per-issue failures.

feat(labeler): implement Linux support verification for server requests

* Added checks for Linux support based on issue content and Steam API data.
* Integrated AI analysis for documentation to assess Linux compatibility.
* Automatically create or remove labels based on support verification results.
* Enhanced feedback for users regarding Linux server support status.