Skip to content

Scrub fleet identifiers from main - #2491

Closed
erikdarlingdata wants to merge 1 commit into
mainfrom
scrub/main-fleet-identifiers
Closed

erikdarlingdata wants to merge 1 commit into
mainfrom
scrub/main-fleet-identifiers

Conversation

@erikdarlingdata

@erikdarlingdata erikdarlingdata commented Aug 21, 2026 •

Copy link
Copy Markdown
Owner

Closed — check-branches refuses it by design ("pull requests to main are only allowed from dev"), and that gate is worth keeping. The branch is retained in case the rename is easier to graft than to re-derive.

Two things worth keeping from the attempt, both fixed elsewhere:

  • A bare-token rename touched 50 files, because apex is domain vocabulary here — the root of a blocking chain. Scoped to whole instance names it is 9 files.
  • A word boundary failed where a hostname really was: in a C# literal following an escape, the n of \n is a word character. The Pin fixture hostnames to synthetic slugs #2490 guard had the same flaw and no longer does.

…fault

The scrub landed on dev only. main is 221 commits behind it and is the branch
anyone browsing this public repo sees first, so the old names are still the
ones on display -- including in Darling/README.md and darling.sample.json,
which are the files a new user actually reads.

Waiting for the release to carry the fix over leaves them up in the meantime,
so this is the same rename applied directly to main.

Scoped to whole instance names rather than bare tokens, because "apex" is
domain vocabulary here -- it is the root of a blocking chain in about fifty
files -- and a token-level rename corrupted every one of them on the first
attempt.

No leading word-boundary assertion either. One name appears as
"servers:\nprod-..." where the n of the escape is a word character, so both
\b and a negative-lookbehind fail at precisely the spot a hostname does
occur. That is the third boundary assumption in this scrub to under-match
and report clean.

Verified the renamed strings are not load-bearing: both peer match patterns
are still contained in the hosts they have to claim, so the disclosure still
agrees in number. Darling.Tests and Lite.Tests both build clean.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
@erikdarlingdata

Copy link
Copy Markdown
Owner Author

Closing this — check-branches refuses it by design:

Pull requests to main are only allowed from dev. Please target the dev branch instead.

That gate is release discipline and worth keeping, so this is the wrong instrument. The scrub is already on dev; main gets it the moment dev merges to main, and the sanctioned path there is a release.

So this is not a patch decision, it is a release-timing one. Tracking in #2492 rather than leaving a permanently red PR open. The branch is kept in case the content is useful as a cherry-pick during the release.

Worth keeping from the attempt, since both were found here and are fixed elsewhere:

  • A bare-token rename corrupted 50 files. apex is domain vocabulary — the root of a blocking chain. Scoped to the instance-name shape it is 9 files, not 58.
  • A word boundary failed where a hostname really was: "servers:\nprod-...", where the n of the escape is a word character. The Pin fixture hostnames to synthetic slugs #2490 guard had the same flaw and no longer does.

Repository owner deleted a comment from claude Bot Aug 22, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant