22 merged PRs have no CHANGELOG entry section at all, so there is nothing to transcribe

[gsdali]

Split out of #1681, which closed the 190 PRs whose entry existed in the PR body and never reached
the file. This is the other half, and it is a different job: these PRs have no
## CHANGELOG entry section in the body, so there is nothing to copy and each entry has to be
written from the merged diff.

check-changelog-transcription.py --verify-transcribed on main at 1a88421e:

transcribed late, nothing to do:            258
entry in the PR body, NOT in the CHANGELOG:   1     <- #1649, see below
PR body has no entry section at all:         32     <- this issue
PR body says None, by design:                27
could not verify:                             1

The 32, and why the real number is 22

Ten of the 32 already have an entry that somebody wrote later. The bucket does not shrink when they
do, because it describes the PR body, not the file, and a body cannot be retroactively given a
section it never had.

PR	Entry in the file?
#907, #1103, #1120	yes, drafted at the time by whoever noticed
#1590, #1591, #1593, #1610, #1612	yes, drafted in #1680
#1666, #1671	yes, drafted in #1680; their bodies were later overwritten with merge-order notes, so nothing in them shows it

The 22 that genuinely have no entry anywhere:

#856, #858, #874, #932, #934, #935, #937, #939, #940, #1091, #1113, #1114, #1115, #1116, #1128,
#1140, #1152, #1351, #1352, #1472, #1520, #1539

Two different answers are correct here

Not all 22 owe an entry. Several look like they owe a No-Changelog: trailer instead, which is the
route the policy already provides and which two merges in this window have used:

• Probably no entry: docs: require a GTest by default on every upstream OCCT PR #856, docs(okf): add the upstream OCCT patch process, start to finish #858, docs(okf): merge upstream-occt-style and patch-process, add format.patch technique #874, docs: sync the context-first policy with the ecosystem canonical #1091, docs: add AGENTS.md for agent workspace guidance #1128 are okf policy and AGENTS.md docs; fix: swift-format violations in IntAna.swift (pre-existing on main) #1539 is
  a swift-format fix to pre-existing violations; chore: retire patch 0032, superseded by OCCT#1505/#1509 + add upstream-check process step #1472 retires patch 0032, which shipped nothing.
  A No-Changelog: reason on the merge is the honest record for these.
• Probably an entry: fix: Add null-wire guards to 19 ShapeAnalysis_Wire bridge functions (closes #1099) #1113 adds null-wire guards to 19 ShapeAnalysis_Wire bridge functions,
  Fix #1087: Edge.adjacentFaces(in:) now returns all adjacent faces for non-manifold edges #1116 makes Edge.adjacentFaces(in:) return all faces for a non-manifold edge, Fix #1073: checkOuterBound area cancellation and partial pcurve set #1140 fixes
  checkOuterBound area cancellation, fix(#1497): OCCTShapeSymmetryAxes honors fractionalTolerance in its existence gate #1520 makes OCCTShapeSymmetryAxes honour
  fractionalTolerance. Those are shipped behaviour changes with no release note.

The split has to be made per PR by reading the merged diff, not guessed from the title.

Method

#1680 did seven of these and is the worked example: read the merge, write the entry as the PR should
have, and say in the transcription PR's body which entries are drafted rather than transcribed,
so the difference stays visible. The policy's "the merger transcribes, they do not draft" is about
not rewriting an author's supplied text; it is not a reason to leave a permanent hole in the release
record when there is no text at all.

Same batching rules as #1681: transcribe-only PRs, and only one open at a time touching
docs/CHANGELOG.md.

The one remaining MISSING row

#1649's entry has been in the file since it merged. The audit reports it missing because the PR
body's heading writes #1399 as a markdown link and the file writes it bare, and
entry_is_present() matches on a substring. Not worth changing the audit for one row, but it is
worth knowing that "1" is the floor, not a gap.

Related

• 191 merged PRs carry a CHANGELOG entry in the PR body that never reached the file #1681, the 190 transcribed in docs(#1681): transcribe the CHANGELOG entries for #1118 through #1235 #1684, docs(#1681): transcribe the CHANGELOG entries for #1236 through #1357 #1685, docs(#1681): transcribe the CHANGELOG entries for #1358 through #1465 #1686 and docs(#1681): transcribe the CHANGELOG entries for #1466 through #1562 #1687.
• The CHANGELOG-on-merge policy has no automated backstop; a forgotten transcription is silent #742, the audit script.
• okf/policies/changelog-on-merge.md, including the No-Changelog: trailer.