[JarryShaw]

• Searched for similar pull requests
• Followed the coding style (make pylint, make mypy, make isort)
• make test passes, and a test case covers the change
• Added a changelog entry — 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?

• fix — bug fixes

---

Description of your pull request and other information

Closes #2120. release_published.var() and release_tiers.var() read os.environb, which does not exist where os.supports_bytes_environ is false, as on Windows. Every conda-dist (windows-latest) leg of Conda Update run 40020510868 therefore died at the first line of release_conda.py build, before conda build ran.

• Fix: a new shared util/_actions.getenvb(name) reads os.environb where it exists and os.fsencode(os.environ[name]) otherwise. Both var()s call it inside their existing except KeyError, so an unset variable still gives <name>: unbound variable and exits 1.
• Rest of the Windows build path: no other POSIX-only stdlib use was found in release_conda.py, release_published.py, _actions.py, leg_summary.py or conda_vendor.py. Not verified: that subprocess finds conda.exe on PATH under bash -el; the next run will show it.
• Test: TestTheEnvironmentWithoutEnvironb in test_conda_windows_leg_2060 removes os.environb and sets supports_bytes_environ = False. It checks the helper and both var()s, that release_conda.main(['build']) gets past the env reads to the conda build call, and that an unset variable still fails the same way. With the fix reverted, 6 checks fail with the CI's AttributeError. The conda, release-script, helper and release-gates tests pass.

[JarryShaw]

GOOD TO GO at f4f24e3b6 (Sonnet review).

• POSIX unchanged: for b'caf\xe9\xff', getenvb returns the same bytes as os.fsencode(os.environ[…]), and an unset name still becomes Unbound.
• util/ clean: it is stdlib-only, in __all__, and has no other os.environb use left. The rest of the build path is portable.
• Tests: they fail with util/ reverted (6 failed). The Windows-leg, release-conda, version-check and helper tests pass.
• UNVERIFIED (reasoning only): that call('conda', …) finds conda.exe on PATH under bash -el. The first win-64 run after this merges will show it.

[github-actions]

Coverage: 91.54% (unit tier, Python 3.15, 8777e5e5a, Unit Tests run success)

Package	Statements	Missed	Branches	Partial	Cover
pcapkit (top level)	104	4	20	2	95.16%
pcapkit/const	19599	880	2618	866	92.12%
pcapkit/corekit	2589	75	816	37	96.48%
pcapkit/dumpkit	282	3	102	3	98.44%
pcapkit/foundation	2959	23	1032	25	98.80%
pcapkit/interface	138	4	52	3	96.32%
pcapkit/protocols	17418	219	4574	185	98.14%
pcapkit/toolkit	869	36	298	10	95.20%
pcapkit/utilities	498	7	144	5	98.13%
pcapkit/vendor	4432	1988	1018	167	51.87%

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

[JarryShaw]

Ready to merge at f4f24e3b6.

• Review: the Sonnet review is GOOD TO GO.
• CI: every blocking check passed (Required checks passed SUCCESS). The 3.8–3.10 KTLO legs carry no warnings.
• Ordering: the tests/project plain-unittest ordering leg passed on a scratch merge into main (82b901b91). It ran 2186 tests with 0 failures.
• On Windows: build-only run 40034513322 (this PR + ci(conda): build the macos-26-intel legs again (#2123) #2124) built win-64 for Python 3.11–3.15. release_conda.py build no longer dies on os.environb. The 3.11–3.14 smoke reads 35 native binaries, none failing. The 3.15 smoke stopped at a separate check bug, CFEP-27's new layout, which fix(conda): read conda-forge's CFEP-27 Windows layout in the smoke check (#2126) #2127 fixes.

This closes #2120. It conflicts with #2127 in one docstring of tests/project/test_conda_windows_leg_2060.py. I'll rebase whichever lands second.

[JarryShaw]

use :pr:/:issue: for the GitHub issue/PR link rendering.

[JarryShaw]

Will do. The workflows.rst row for util/_actions.py becomes :issue:`2047` and :issue:`2120` , in place of the hand-written links. A fresh review follows the push. The rest of the repo's hand-written links go to #1103's next round.

[JarryShaw]

use directives in Sphinx docs only. use #NNN format in plain comments and texts so that GitHub can auto render them.

[JarryShaw]

Noted. The roles go only where Sphinx renders: .rst files, and the docstrings and #: comments that autodoc renders. Plain comments, util//tests/ docstrings, commit messages and GitHub text keep #NNN. This PR's other mentions of #2120 are all in util/ and tests/, so they stay as they are. #1103's reminder already scopes the sweep the same way.

[JarryShaw]

GOOD TO GO at 3ce5b480e (Sonnet delta review).

• Rebase: onto main 7904f615b. The range-diff against f4f24e3b6 shows only the workflows.rst row for util/_actions.py, which now reads :issue:`2047` and :issue:`2120` .
• Docs: the issue extlink is defined in conf.py, and the list-table cell's reST is valid.
• Scope: the PR's other #2120 mentions are in util/ and tests/, which Sphinx does not render, so they keep #NNN.
• Tests: tests/project/test_conda_windows_leg_2060.py gives 26 passed.

I'll post the ready note when CI finishes on the new head.

[JarryShaw]

Rebased 3ce5b480e → 8777e5e5a onto main 23a815b84, after #2127 landed.

• The conflict: both PRs extended the module docstring of tests/project/test_conda_windows_leg_2060.py. The resolution keeps both: fix(conda): read conda-forge's CFEP-27 Windows layout in the smoke check (#2126) #2127's CFEP-27 sentence and this PR's TestTheEnvironmentWithoutEnvironb bullet.
• Everything else is unchanged: the per-file patch-ids of workflows.rst, util/_actions.py, util/release_published.py and util/release_tiers.py are identical to the old head's. The test file differs only in that docstring context, and gives 28 passed.
• Old head's CI: it had not finished, with 40 checks still queued or running, so CI has to run on the new head before this is ready.

[JarryShaw]

GOOD TO GO at 8777e5e5a (Sonnet rebase check). The range-diff against 3ce5b480e differs only in the docstring resolution, which keeps both sides. The diff touches only this PR's five files, and tests/project/test_conda_windows_leg_2060.py gives 28 passed.

[JarryShaw]

Ready to merge at 8777e5e5a.

• Review: GOOD TO GO, from the Sonnet delta review of the :issue: change and the Sonnet rebase check.
• CI: every blocking check passed (Required checks passed SUCCESS). The 3.8–3.10 KTLO legs carry no warnings.
• Ordering: the tests/project leg passed (2186 tests, 0 failures) at f4f24e3b6. Since then the per-file patch-ids of workflows.rst and the util/ files are identical, and the test file differs only in the docstring resolution. The head sits directly on main 23a815b84.

This closes #2120. Once it lands, #2060 phase 2 needs one Conda Update run on main with clean Windows legs.