[JarryShaw]

Please follow the guide below

• You will be asked some questions, please read them carefully and answer honestly

• Put an x into all the boxes [ ] relevant to your pull request (like that [x])

• Use Preview tab to see how your pull request will actually look like

• Searched for similar pull requests

• Followed the coding style (make pylint, make mypy, make isort) — not run: all three lint pcapkit, and no pcapkit/ file changed; the new test is isort -l100 clean

• make test passes, and a test case covers the change — ran the affected test files, not the whole suite

• Added a changelog entry under docs/source/changelog/ and regenerated CHANGELOG.md, if the change is user-visible — N/A — centralised in docs(changelog): shared 1.5.0 changelog — long-lived, merges last (#610, #616, #617, #618, #620) #657

What is the purpose of your pull request?

Tick the commit type your subject line carries.

• fix — corrects a defect
• feat — adds a feature
• perf — changes performance, not behaviour
• refactor — changes neither behaviour nor performance
• test — tests only
• docs — documentation only
• ci — workflows or build tooling
• release — bumps the version or rolls up a distribution
• chore — anything else

---

Description of your pull request and other information

Cause. At v1.21.0, conda-dist's allowUpdates: true calls Update a release before it uploads (src/Action.ts:107-120). That call sends name, body, prerelease and draft (src/Releases.ts:162-174). Unset, those are the tag (src/Inputs.ts:128-135), "" (:70-82) and false (:84-92, :210-240). The issue's :103-118 is createdReleaseName and commit; the tag fallback is :128-135.

Fix. The step now sets the four omit*DuringUpdate inputs. Each one makes its getter return undefined (:210-252). Octokit deletes undefined keys before sending (dist/index.js:38369-38388), and Update a release requires no body parameter (REST docs). omitBodyDuringUpdate also stops generated notes on update (src/Action.ts:192-194). releasing.rst is updated to match.

Other inputs. An empty discussionCategory becomes undefined (Inputs.ts:120-126), so it is dropped. makeLatest is never empty: its default, legacy (action.yml:58-61), is sent on every update (job 107491856892 logs makeLatest: legacy), and the action has no omit input for it. I left it alone.

Evidence. 33 of the 37 releases with bump (_N, N ≥ 1) conda assets have a name equal to their tag and a 0-character body, v1.5.0a1, v1.4.1.post2, v1.4.1 and v1.3.5.post42 among them. v1.5.0b4's bump upload succeeded on 2026-09-24, yet it now has its name and 2,487 characters of notes, and its updated_at is 2026-09-29. v1.5.0a1 still reads prerelease: true. So the prerelease reset comes from reading the source; I did not observe it.

Tests. The new test_conda_dist_keeps_release_notes_1657.py pins the four inputs and the step's exact input set, checked against both #1634's scanner and yaml.safe_load. With unittest it reports Ran 4 tests … OK. With origin/main's cron-conda.yml swapped in it reports FAILED (failures=6). The other tests that read these files also pass (#1634, release gates, unit-tests queue, doc claims, workflow and #1633 tests).

Second commit. Once a draft stays a draft, the release check's reason for skipping one is stale. Its comment and ::notice now say the upload would retarget the draft at main (src/Action.ts:85-92, :107-120) rather than publish it, and releasing.rst and test_release_tags_the_run_sha_1634.py say the same. The skip itself is unchanged. The updated draft pin fails on the previous text (1 failed, 65 passed).

Third commit. The second said a re-run of Create Release would reuse the retargeted draft, target and all. It would not: action-gh-release 3.0.2 finds drafts (src/github.ts:832-866) and overwrites a differing non-empty target_commitish (:636-646), which create-release.yml sets to github.sha on workflow_run. The comment, releasing.rst and the #1634 docstring now give tagging main's tip to the missing-release path only, and call skipping a draft a precaution. Wording only; the notices and their pins are unchanged.

Not done. No notes or assets are restored.

Closes #1657

[JarryShaw]

Verdict: GOOD TO GO at edcf34396. This took three Sonnet-reviewed rounds; rounds 2 and 3 corrected wording only. I re-ran the 1657, 1634 and release-gate tests: 4, 66 and 120 passed.

The fix. conda-dist's ncipollo/release-action@v1.21.0 step now sets omitBodyDuringUpdate, omitDraftDuringUpdate, omitNameDuringUpdate and omitPrereleaseDuringUpdate. An upload therefore leaves the release's name, notes, prerelease flag and draft state alone (src/Inputs.ts:210-252). Live data shows the damage this stops: 33 of the 37 releases with bump assets have their name reset to the tag and an empty body.

Rounds 2 and 3. The release check's reasons now match the pinned sources. A missing release would be created, and tagged, at main's tip. A draft would be retargeted at commit; Create Release's workflow_run then resets it to its own sha (softprops v3.0.2 src/github.ts:636-646). So skipping a draft is a precaution only.

No notes or assets are restored; that is still your call on #1657.

[github-actions]

Coverage: 89.46% (unit tier, Python 3.14, edcf34396, Unit Tests run success)

Package	Statements	Missed	Branches	Partial	Cover
pcapkit (top level)	104	3	20	3	95.16%
pcapkit/const	18858	954	2366	868	91.06%
pcapkit/corekit	2366	67	760	35	96.48%
pcapkit/dumpkit	275	3	100	3	98.40%
pcapkit/foundation	3031	152	1074	35	94.23%
pcapkit/interface	138	7	52	5	93.68%
pcapkit/protocols	17166	226	4580	190	98.07%
pcapkit/toolkit	1350	275	558	6	77.20%
pcapkit/utilities	494	4	146	4	98.75%
pcapkit/vendor	4409	2342	1006	157	43.25%

Per-file detail: the coverage-html artifact of this run.