Skip to content

fix(tests): wrap Path.expanduser() to never raise on Windows runner#1276

Merged
danielmeppiel merged 2 commits into
mainfrom
fix/windows-expanduser
May 11, 2026
Merged

fix(tests): wrap Path.expanduser() to never raise on Windows runner#1276
danielmeppiel merged 2 commits into
mainfrom
fix/windows-expanduser

Conversation

@danielmeppiel
Copy link
Copy Markdown
Collaborator

@danielmeppiel danielmeppiel commented May 11, 2026

Problem

The v0.13.0 retag CI still red-marked Windows on two tests after #1272:

FAILED tests/unit/install/test_user_scope_rejection_reason.py::test_user_scope_accepts_tilde_local_path[~/pkg]
FAILED tests/unit/install/test_user_scope_rejection_reason.py::test_user_scope_accepts_tilde_local_path[~/sub/pkg]
       raise RuntimeError("Could not determine home directory.")
E      RuntimeError: Could not determine home directory.
C:\hostedtoolcache\windows\Python\3.12.10\x64\Lib\pathlib.py:1406: RuntimeError

Different code path than #1272: production code in install.package_resolution.user_scope_rejection_reason calls Path('~/pkg').expanduser() to detect absoluteness. ntpath.expanduser raises RuntimeError when both USERPROFILE and HOMEPATH are absent — the windows-2025-vs2026 runner state.

Fix

tests/conftest.py (root, loaded by every xdist worker before any test): wrap Path.expanduser. On RuntimeError, fall back to _TMP_HOME (the same hermetic dir Path.home() uses). Expanded path is still absolute so production code's .is_absolute() check still behaves correctly.

Validation

  • New regression-trap test in tests/unit/test_path_home_override.py (clears HOME/USERPROFILE/HOMEDRIVE/HOMEPATH, asserts Path('~/pkg').expanduser() returns an absolute path without raising).
  • The two previously-failing tests now pass locally.
  • Lint clean.

Why ship now

Required for v0.13.0 retag to go green on Windows.

fix(tests): wrap Path.expanduser() to never raise on Windows runner
b9b5e66 
Companion to #1272. After the Path.home() override landed, the
windows-2025-vs2026 runner still red-marked two tests:

  tests/unit/install/test_user_scope_rejection_reason.py
    ::test_user_scope_accepts_tilde_local_path[~/pkg]
    ::test_user_scope_accepts_tilde_local_path[~/sub/pkg]

These hit a different code path: production code in
install.package_resolution.user_scope_rejection_reason calls
Path('~/pkg').expanduser() to detect that the path is absolute.
ntpath.expanduser raises RuntimeError('Could not determine home
directory.') when both USERPROFILE and HOMEPATH are absent --
exactly the windows-2025-vs2026 worker env state.

Fix in tests/conftest.py (root, loaded by every xdist worker before
any test):

  - Wrap Path.expanduser. On RuntimeError, fall back to _TMP_HOME
    (the same hermetic dir Path.home() uses). The expanded path is
    still absolute, so production code's .is_absolute() check keeps
    behaving correctly.

Regression trap added in tests/unit/test_path_home_override.py: clears
HOME/USERPROFILE/HOMEDRIVE/HOMEPATH and asserts Path('~/pkg').expanduser()
returns an absolute path without raising.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Copilot AI review requested due to automatic review settings May 11, 2026 13:25
chore: correct PR number in CHANGELOG entry
03d6a1b 
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
@danielmeppiel danielmeppiel merged commit 9216516 into main May 11, 2026
9 checks passed
@danielmeppiel danielmeppiel deleted the fix/windows-expanduser branch May 11, 2026 13:47
@danielmeppiel danielmeppiel review requested due to automatic review settings May 11, 2026 13:50
@danielmeppiel danielmeppiel mentioned this pull request May 11, 2026
Open
danielmeppiel pushed a commit that referenced this pull request May 18, 2026
docs(changelog): tighten v0.14.0 entries; add post-cut PRs
9003a3b 
- One concise line per PR answering 'so what?' for end users
- Add 5 missing entries: #1376 (perf resolver), #1373 (shared/apm.md
  matrix secret-stripping), #1246 (install.ps1 GHES env vars), #1255
  (warn missing apm.yml), #1248 (extends:org unmanaged_files)
- Drop internal/CI/test-infra entries (#1270, #1271, #1272, #1274,
  #1276, #1291, #1360 refactor)
- Consolidate three #605 lines and four #1317 lines into one entry
  per PR where appropriate
- Promote MCP Registry v0.1 to a dedicated Breaking section

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
danielmeppiel added a commit that referenced this pull request May 18, 2026
@danielmeppiel
chore: cut 0.14.0 (#1372)
9d1161b 
* chore: cut 0.14.0

Renames the [Unreleased] block in CHANGELOG.md to [0.14.0] - 2026-05-18
and bumps the package version from 0.13.0 to 0.14.0 in pyproject.toml
(and uv.lock by regeneration).

0.14.0 ships the producer-experience epic (#1348) on the CLI side --
notably:

- apm pack --check-versions / --check-clean (#1365), the release gates
  consumed by apm-action mode: release.
- apm plugin init (#1370), the noun-verb successor to apm init --plugin.
- apm pack multi-format outputs (--marketplace, --marketplace-path,
  --json, marketplace.outputs map form) (#1317).
- New producer docs corpus (repo-shapes / releasing-from-any-ci /
  versioning-strategies) (#1370).
- Breaking: MCP registry client adopts the official v0.1 spec; self-
  hosted registries must serve /v0.1/ paths (#1337).

Plus the deprecations and fixes already listed in the moved block.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* docs(changelog): tighten v0.14.0 entries; add post-cut PRs

- One concise line per PR answering 'so what?' for end users
- Add 5 missing entries: #1376 (perf resolver), #1373 (shared/apm.md
  matrix secret-stripping), #1246 (install.ps1 GHES env vars), #1255
  (warn missing apm.yml), #1248 (extends:org unmanaged_files)
- Drop internal/CI/test-infra entries (#1270, #1271, #1272, #1274,
  #1276, #1291, #1360 refactor)
- Consolidate three #605 lines and four #1317 lines into one entry
  per PR where appropriate
- Promote MCP Registry v0.1 to a dedicated Breaking section

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* docs(changelog): add #1377 Bitbucket DC tilde fix

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

---------

Co-authored-by: Daniel Meppiel <copilot-rework@github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
sergio-sisternes-epam pushed a commit that referenced this pull request May 19, 2026
@danielmeppiel
fix(tests): wrap Path.expanduser() to never raise on Windows runner (#…
668d215 
…1276)

* fix(tests): wrap Path.expanduser() to never raise on Windows runner

Companion to #1272. After the Path.home() override landed, the
windows-2025-vs2026 runner still red-marked two tests:

  tests/unit/install/test_user_scope_rejection_reason.py
    ::test_user_scope_accepts_tilde_local_path[~/pkg]
    ::test_user_scope_accepts_tilde_local_path[~/sub/pkg]

These hit a different code path: production code in
install.package_resolution.user_scope_rejection_reason calls
Path('~/pkg').expanduser() to detect that the path is absolute.
ntpath.expanduser raises RuntimeError('Could not determine home
directory.') when both USERPROFILE and HOMEPATH are absent --
exactly the windows-2025-vs2026 worker env state.

Fix in tests/conftest.py (root, loaded by every xdist worker before
any test):

  - Wrap Path.expanduser. On RuntimeError, fall back to _TMP_HOME
    (the same hermetic dir Path.home() uses). The expanded path is
    still absolute, so production code's .is_absolute() check keeps
    behaving correctly.

Regression trap added in tests/unit/test_path_home_override.py: clears
HOME/USERPROFILE/HOMEDRIVE/HOMEPATH and asserts Path('~/pkg').expanduser()
returns an absolute path without raising.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* chore: correct PR number in CHANGELOG entry

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

---------

Co-authored-by: Daniel Meppiel <copilot-rework@github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
sergio-sisternes-epam pushed a commit that referenced this pull request May 19, 2026
@danielmeppiel
chore: cut 0.14.0 (#1372)
0678cc8 
* chore: cut 0.14.0

Renames the [Unreleased] block in CHANGELOG.md to [0.14.0] - 2026-05-18
and bumps the package version from 0.13.0 to 0.14.0 in pyproject.toml
(and uv.lock by regeneration).

0.14.0 ships the producer-experience epic (#1348) on the CLI side --
notably:

- apm pack --check-versions / --check-clean (#1365), the release gates
  consumed by apm-action mode: release.
- apm plugin init (#1370), the noun-verb successor to apm init --plugin.
- apm pack multi-format outputs (--marketplace, --marketplace-path,
  --json, marketplace.outputs map form) (#1317).
- New producer docs corpus (repo-shapes / releasing-from-any-ci /
  versioning-strategies) (#1370).
- Breaking: MCP registry client adopts the official v0.1 spec; self-
  hosted registries must serve /v0.1/ paths (#1337).

Plus the deprecations and fixes already listed in the moved block.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* docs(changelog): tighten v0.14.0 entries; add post-cut PRs

- One concise line per PR answering 'so what?' for end users
- Add 5 missing entries: #1376 (perf resolver), #1373 (shared/apm.md
  matrix secret-stripping), #1246 (install.ps1 GHES env vars), #1255
  (warn missing apm.yml), #1248 (extends:org unmanaged_files)
- Drop internal/CI/test-infra entries (#1270, #1271, #1272, #1274,
  #1276, #1291, #1360 refactor)
- Consolidate three #605 lines and four #1317 lines into one entry
  per PR where appropriate
- Promote MCP Registry v0.1 to a dedicated Breaking section

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* docs(changelog): add #1377 Bitbucket DC tilde fix

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

---------

Co-authored-by: Daniel Meppiel <copilot-rework@github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@sergio-sisternes-epam sergio-sisternes-epam Awaiting requested review from sergio-sisternes-epam sergio-sisternes-epam is a code owner

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

1 participant

@danielmeppiel