[JarryShaw]

Please follow the guide below

• Searched for similar pull requests
• Followed the coding style (make pylint, make mypy, make isort) -- tests/project/test_isort_clean.py passes; no other lint run (advisory in CI)
• make test passes, and a test case covers the change -- tests/project/test_conda_dev_release_gate_2458.py (new) and tests/project/test_cron_reports_2030.py; tests/project/test_release_gates.py and every other conda-* test file still pass unmodified
• Added a changelog entry under docs/source/changelog/ and regenerated CHANGELOG.md, if the change is user-visible -- not applicable here; out of scope for this PR

What is the purpose of your pull request?

• ci -- workflows or build tooling

---

Description of your pull request and other information

Closes #2458.

.github/workflows/create-release.yml's publishers are being gated on a dev-release version by #2452's (proposed, not yet merged) owner ruling: __version__ carrying a PEP 440 dev-release suffix (X.Y.Z.devN) is a work-in-progress signal, and that workflow's workflow_run trigger means a scheduled Vendor Update can reach it and publish that version with no human in the path.

.github/workflows/cron-conda.yml has its own independent route to publishing: conda-tag pushes a conda-<version>+<build> tag and conda-dist uploads to anaconda.org, both whenever mode's publishes output is 'true' and conda-update's CONDA_CHANGED is 'true' -- reachable the same way from its own schedule/workflow_run triggers, and until now with no dev-release check of its own. This PR:

• adds PCAPKIT_DEV_RELEASE to conda-update's Get Version step, from packaging.version.Version(...).is_devrelease (deliberately not is_prerelease, which a shipped alpha/beta/rc also satisfies).
• gates conda-tag's if: on it as a top-level conjunct, and conda-dist's inside its publishing branch only, so a build-only manual run still builds and smoke-tests a dev version without uploading it.
• makes the discussion job's "Post the report to Discussions" step explain a pushed-but-withheld dev-release build ("Tag: skipped, and Upload: skipped -- ... withheld both deliberately") instead of silently dropping the Tag/Upload lines, so "nothing was pushed" and "a dev build was pushed but withheld" read differently.

Test plan

• tests/project/test_conda_dev_release_gate_2458.py (new): the if: conditions, behavioural rows through job_runs/conda_context, and the real Get Version script against a 1.5.0b14 stub vs dev-suffixed stubs.
• tests/project/test_cron_reports_2030.py: the dev-release explanation line, plus rows pinning it silent whenever CONDA_CHANGED or DEV_RELEASE is not literally 'true' (including an empty DEV_RELEASE).
• Both test files together: 43 passed, 105 subtests passed.
• actionlint 1.7.12 (shellcheck 0.10.0): zero findings on cron-conda.yml.

[JarryShaw]

Cross-review (different model, adversarial) came back NEEDS CHANGES. The gate logic itself is verified correct (288-state evaluation, zero regressions), but:

1. The conda run report goes silent/misleading on a gated dev-version run — mirrors ci: gate create-release.yml's publishers on a dev-release version #2452's own release_status finding, just on this workflow's own report script.
2. Two narrowed-predicate mutations (adding and Version(v).pre is None / and not is_postrelease) survive all 13 tests — the suite needs a .postN.devM and a rcN.devM case in the real-script test, not just the hand-picked forms.
3. Some comments/docstrings describe ci: gate create-release.yml's publishers on a dev-release version #2452's gate as already present on main — it's still open at review: needs-changes, so those need rewording to describe it as a sibling proposal, not an existing fact.

Dispatching a fix to this branch now.

[JarryShaw]

Fixed and verified directly (new head 27f36f901): the discussion-report step now distinguishes "nothing pushed" from "a dev-release build was pushed but tag/upload withheld on purpose", added the 2 missing test cases that closed the narrowed-predicate gap, and reworded the stale #2452-as-fact references. 55 tests/118 subtests passing, actionlint clean.

[JarryShaw]

Cross-review (different model, adversarial), round 3, came back NEEDS CHANGES. The report fix itself is verified correct for the real scenarios — but 3 mutations survive because the test suite only checks the "should explain" case, never the "should stay quiet" case:

1. A dev version whose build wasn't pushed (CONDA_CHANGED=false) should NOT get the tag/upload explanation line, but nothing tests this — surviving mutation: dropping the CONDA_CHANGED guard entirely.
   2 & 3. Two more survive on the DEV_RELEASE emptiness check ([ -n "$DEV_RELEASE" ] and != false) for the same reason — no negative-case coverage.

Also: the commit message/PR body claim "55 passed, 118 subtests" — measured 42/102 for this round's actual diff (the higher count looks like it picked up a stale copy of an earlier round's test file). The PR body is also out of date (still describes an earlier round's scope).

Dispatching a fix now.

[JarryShaw]

Fixed and verified directly (new head 609291943): added the 3 "should stay quiet" test rows, each confirmed to fail against the specific mutation it closes (verified by editing the workflow's elif line in place, confirming failure, then restoring). Corrected the inflated test-count claim — real count is 43 passed/105 subtests, not 55/118 (the earlier number picked up a stale duplicate test file). PR body rewritten to describe the current state accurately.

[JarryShaw]

Round 5: wording-only fix needed — commit claims mypy clean, but it isn't on the 2458 test file (15 attr-defined + 3 unused-ignore, pre-existing pattern, not CI-gated). Fixing the claim.

[JarryShaw]

Wording-only fix confirmed (zero file diff). Good to go.