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
fix: promote-flow dev version rebase and e2e entry-setup gate (#2040)
* fix: base reset dev version on the next stable patch
reset_version computed new_base = max(stable_base, dev_base), so after
every normal promote the dev base equaled stable's — and PEP 440 sorts
X.Y.Z.devN BEHIND X.Y.Z, so the dev channel advertised a version reading
older than the stable carrying the same code. AGENTS.md has always
documented the next-stable-patch rule; nothing enforced it, so the drift
was silent (Patch76's review finding on #2037). The base is now
max(next_patch(stable), dev_base) with the counter restarting on a base
jump, and the test suite pins the sorts-ahead-of-stable invariant.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017iTWsCSMPqSHm7Po8yGzfk
* fix: gate e2e bring-up on all config entries finishing setup
The unloaded-entry diagnostic named the recurring
test_include_options_flag failure: the seeded HACS entry is still
not_loaded when CoreState.RUNNING trips (RUNNING only means setup was
dispatched), so the OptionsFlow probe sees all-empty options while a
sibling test probing the same entry seconds later passes. Add the
follow-up gate the drift warning has been calling justified: a bounded
wait for entries_loaded == entries_total, entered only when the
trip-time snapshot caught a straggler, logging loudly on timeout.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017iTWsCSMPqSHm7Po8yGzfk
* feat: auto-dispatch the dev reset when a promote PR merges
The promotion flow ended at the draft promote PR; rebasing the dev
version line onto the new stable was an undocumented manual dispatch.
A post-promote workflow now fires Reset Dev from Stable on promote-PR
merge (human-merged, so the event triggers; workflow_dispatch is
GITHUB_TOKEN-permitted), and the promotion doc gains the missing step.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017iTWsCSMPqSHm7Po8yGzfk
* fix: rebase the dev version line inside the promote PR itself
Replace the just-added post-promote dispatcher: instead of spawning a
reset PR after every promote merge, the promote transform now rewrites
the dev flavor's two version files (rebase_dev_version — next stable
patch + .dev1) in the same PR, the workflow stages them, and the doc's
step 6 describes it. Reset Dev from Stable remains a manual tool for
the stable-hotfix backport case.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017iTWsCSMPqSHm7Po8yGzfk
* chore(addon-dev): rebase dev version line onto stable 2.1.0 (2.1.1.dev1)
The 2.1.0 promotion (#2037) predates the in-PR rebase this branch adds,
so dev still advertised 2.0.5.dev3 — sorting behind the stable carrying
the same code. One-time application of rebase_dev_version; future
promotions carry this inside the promote PR itself.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017iTWsCSMPqSHm7Po8yGzfk
* fix: enter the entries_loaded gate when the trip snapshot is unavailable
A failed trip-time snapshot (snapshot_ok=False) returns an empty
unloaded string, which skipped the follow-up gate while entries could
still be loading — preserving the race behind a transient HTTP/JSON
hiccup. The skip now requires positive all-loaded confirmation; the
gate's loop polls through snapshot failures and stays bounded.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017iTWsCSMPqSHm7Po8yGzfk
---------
Co-authored-by: kingpanther13 <kingpanther13@users.noreply.github.com>
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
0 commit comments