Skip to content

Re-take the bundle document against what the build writes now - #212

Closed
iderex wants to merge 1 commit into
mainfrom
documentation/the-bundle-document-is-re-taken-against-what-the-build-writes
Closed

Re-take the bundle document against what the build writes now#212
iderex wants to merge 1 commit into
mainfrom
documentation/the-bundle-document-is-re-taken-against-what-the-build-writes

Conversation

@iderex

@iderex iderex commented Aug 28, 2026

Copy link
Copy Markdown
Contributor

Closed, and replaced by #213 with the same bytes

This branch was pushed without a Signed-off-by: trailer on its one commit, and
the DCO check reds every non-merge commit that carries none:

gh api repos/Flowfin/site/commits/96e9319dde6b54ad0e6c6dcafbcb700a356b85b2/check-runs \
  --jq '.check_runs[] | select(.name=="DCO sign-off") | {name,conclusion}'
{"conclusion":"failure","name":"DCO sign-off"}

Run 2026-08-28. That trailer is part of the commit, so adding it to a commit
already pushed means rewriting it. This repository does not force-push, so the
correction goes out under a new name and this branch is abandoned instead.

The replacement carries the same bytes, compared rather than asserted:

git show 96e9319 | git patch-id --stable
d604409ffac8664c5419bf0f353d5088c21d8827 96e9319dde6b54ad0e6c6dcafbcb700a356b85b2
git show 444c77c | git patch-id --stable
d604409ffac8664c5419bf0f353d5088c21d8827 444c77c1f2317ad7e3d0b0f595371d2e8dd008ef

Both run 2026-08-28. The whole of what this change is and what was run is in the
body of #213. Nothing about it changed here.

Six of the outputs pasted into docs/serving-the-bundle.md on 2026-08-16 no
longer reproduce, and the largest of them tells an operator the wrong thing
about what they downloaded. The archive listing named seven files; the build
writes twenty-two, so the install page, the design system page, the icon and
twelve plugin pages were missing from the list somebody checks their unpacked
bundle against. The invariant rows read 4 and 7 files and read 18 and 22, the
reproduce leg read 7 files and reads 22, the two pack digests have moved, the
not-found page has gained a reference, and the sitemap has gained fourteen
entries.

What prevents the same drift where it can be prevented: the two outputs that
were enumerations are replaced by the command and a count, which is the repair
the threat model already took for a hand count. A list of one entry per page
goes stale on the day the next page lands and nothing reddens, so this document
hands the reader the command instead of the answer. The rest are numbers and
are re-taken with the date they were taken on.

The three pastes that still reproduce are left alone: the bill of materials
excerpt, the not-found file location and the empty release listing were re-run
and are unmoved.

The last section says plainly that nothing in this tree holds this document to
the output it describes, so a reader knows the figures are somebody's readings
rather than something a run keeps honest.
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