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
{{ message }}
Repository navigation
Commit 9435bb7
Browse filesBrowse the repository at this point in the historyBrowse files
feat(coil): cut the sync automation over to coil names (#71)
workflows t3x-{ci,upstream-sync,sync-resolve,weekly-verify}.yml -> coil-*
workflow names "t3x fork CI" etc -> "coil fork CI"
label t3x-sync -> coil-sync
branches t3x/sync-* -> coil/sync-*
tags t3x/{pre-sync,last-good}-* -> coil/*
launchd dev.t3x.autobuild -> dev.coil.autobuild
These strings are matched by workflows that are already running, so this is a cutover, not a
find-and-replace, and two of them keep accepting the old value for one cycle:
- coil-ci.yml's push trigger matches BOTH coil/sync-** and t3x/sync-**. A sync branch pushed
before this lands is still open after it, and it is exactly the branch that cannot fall back to
pull_request — a conflicting sync PR has no merge ref, which is the whole reason the push
trigger exists. Dropping the old prefix early ungates precisely the branches that most need it.
- coil-sync-resolve.yml's label gate accepts coil-sync OR t3x-sync, so an escalation issue opened
before the rename can still be resolved by comment afterwards.
The LaunchAgent is the part that fails silently. ~/Library/LaunchAgents/dev.t3x.autobuild.plist
is installed outside this repo and names the pre-rename script path absolutely, so moving
scripts/ leaves launchd invoking something that no longer exists — no error anywhere the user
looks, the nightly build just stops. Same failure class as #41, which cost 103 minutes. It has to
be unloaded and reinstalled by hand; the PR body carries the commands.
The T3X_AUTOBUILD_* variable names inside that plist are deliberately NOT renamed. The plist is
regenerated anyway because the path moved, so a regenerated plist and an unchanged script agree by
construction; renaming the variables too would open a window where a hand-edited plist sets names
the script no longer reads and silently falls back to defaults for the build worktree. The state
files it writes do move, because losing them costs one redundant rebuild.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
FIX='▶ **To fix:** comment `@claude resolve` on this issue (or run the *t3x sync resolve (agent)* workflow). An agent replays the rebase, resolves the conflicts, runs typecheck/lint/test, and opens a PR for review — nothing is pushed to `main`. Runbook: `docs/coil/sync-agent-runbook.md`.'
149
+
FIX='▶ **To fix:** comment `@claude resolve` on this issue (or run the *coil sync resolve (agent)* workflow). An agent replays the rebase, resolves the conflicts, runs typecheck/lint/test, and opens a PR for review — nothing is pushed to `main`. Runbook: `docs/coil/sync-agent-runbook.md`.'
150
150
151
151
if [[ -n "$PUSH_CODE" && "$PUSH_CODE" != "0" ]]; then
152
152
# Rebase and verify were GREEN — the status JSON says so, which is exactly why both
@@ -160,10 +160,10 @@ jobs:
160
160
fi
161
161
EXTRA="$(printf '\n\n**The rebase verified clean — only the push failed.** `origin/main` is unchanged, nothing was force-landed, and open PRs are untouched. Re-running this workflow replays the whole sync safely.\n\n%s\n\n<details><summary>git push output</summary>\n\n```\n%s\n```\n\n</details>' \
162
162
"$HINT" "$PUSH_ERROR")"
163
-
FIX='▶ **To fix:** this is a credential problem, not a conflict — do **not** run the resolver agent. Fix the push credential above, then re-run *t3x upstream sync (daily)*. Runbook: `docs/coil/sync-agent-runbook.md`.'
163
+
FIX='▶ **To fix:** this is a credential problem, not a conflict — do **not** run the resolver agent. Fix the push credential above, then re-run *coil upstream sync (daily)*. Runbook: `docs/coil/sync-agent-runbook.md`.'
164
164
fi
165
165
166
-
TITLE="[t3x-sync] daily rebase needs attention ($RESULT)"
166
+
TITLE="[coil-sync] daily rebase needs attention ($RESULT)"
167
167
168
168
BODY="$(printf 'Automated daily upstream sync could not complete.\n\n**Result:** `%s`\n**Recovery tag:** `%s`\n**Run:** %s%s\n\n<details><summary>status JSON</summary>\n\n```json\n%s\n```\n\n</details>\n\n---\n\n%s' \
BODY="$(printf 'Weekly full build failed.\n\n**kind:** `weekly-build`\n**failed steps:**%s\n**run:** %s\n\nThis is advisory — it does not block the daily rebase. A green typecheck/lint/test can still hide a build regression; investigate when convenient.' "$FAILED_STEPS" "$RUN_URL")"
73
73
74
74
# See the daily workflow for why we decouple label attachment from issue creation:
@@ -77,7 +77,7 @@ jobs:
77
77
# fall back to the title marker; create without --label and attach afterwards.
78
78
EXISTING="$(gh issue list --label "$SYNC_LABEL" --state open --json number --jq '.[0].number // empty' 2>/dev/null || true)"
79
79
if [[ -z "$EXISTING" ]]; then
80
-
EXISTING="$(gh issue list --search 't3x-sync in:title' --state open --json number --jq '.[0].number // empty' 2>/dev/null || true)"
80
+
EXISTING="$(gh issue list --search 'coil-sync in:title' --state open --json number --jq '.[0].number // empty' 2>/dev/null || true)"
0 commit comments