Fixes a pytest-xdist regression after
762bb61562 (not yet released).
pytest-xdist patches sys.stderr with an object which doesn't have
`encoding`. Strictly speaking, this should be fixed there (or more
precisely, in execnet), but it will drop support for older versions
which don't want.
But in any case, the fix turns out to simplify the code, using FD
support added in Python 3.5, so it's good anyway!
Refs: https://github.com/pytest-dev/pytest-xdist/pull/900
* faulthandler: fix an EncodingWarning
* _py/path: tiny change to `ensure` to silence EncodingWarning
We're not supposed to diverge here, but make this change to fix an
unavoidable EncodingWarning that is otherwise raised in pytest's test
suite. The behavior should be exactly the same besides the warning,
hopefully that won't cause confusion.
Since `Traceback.getcrashentry` takes the `ExceptionInfo`, it is not
really independent of it and is in the wrong layer. Prevent nonsensical
mistakes by inlining it.
TracebackEntry needs the excinfo for the `__tracebackhide__ = callback`
functionality, where `callback` accepts the excinfo.
Currently it achieves this by storing a weakref to the excinfo which
created it. I think this is not great, mixing layers and bloating the
objects.
Instead, have `ishidden` (and transitively, `Traceback.filter()`) take
the excinfo as a parameter.
Closes#10831.
This fixes a small bug where running tests that contained
`pytest.fail(pytrace=False)` with the `--tb=line` flag set results in
an output of "None" in the Failures section of the output, and adds
a test to ensure the behavior is correct.
The problem is that we would loop over all directories of the basetemp directory searching for dead symlinks, for each test, which would compound over the test session run.
Doing the cleanup just once, at the end of the session, fixes the problem.
Fix#10896
By "empty traceback" I mean a traceback all of whose entries have been
filtered/cut/pruned out.
Currently, if an empty traceback needs to be repr'ed, the last entry
before the filtering is used instead (added in
accd962c9f).
Showing a hidden frame is not so good IMO. This commit does the
following instead:
1. Shows details of the exception.
2. Shows a message about how the full trace can be seen.
Example:
```
_____________ test _____________
E ZeroDivisionError: division by zero
All traceback entries are hidden. Pass `--full-trace` to see hidden and internal frames.
```
Also handles `--tb=native`, though there the `--full-trace` bit is not
shown.
This commit contains some pieces from
431ec6d34e (which has been reverted).
Helps towards fixing issue # 1904.
Co-authored-by: Felix Hofstätter <Felhof1@hotmail.com>
The old-style `sys.exc_info()` triplet is redundant nowadays with
`(type(exc), exc, exc.__traceback__)`, and is beginning to get
soft-deprecated in Python 3.12.
Add a nicer API to ExceptionInfo which takes just the exc instead of the
triplet. There are already a few internal uses which benefit.
Fix#10875
Without this, fails with
```
...
E AttributeError: 'TestCaseFunction' object has no attribute 'addDuration'
...
E RuntimeWarning: TestResult has no addDuration method
```
pytest could crash given pathological AST position attributes, which shouldn't happen when testing real Python code, but could happen when testing AST produced by e.g. Hylang.
Another example of the failure is in the nightly CI for the JAX project: https://github.com/google/jax/actions/runs/4607513902/jobs/8142126075
Co-authored-by: Bruno Oliveira <nicoddemus@gmail.com>
Co-authored-by: Jake VanderPlas <jakevdp@google.com>
Nowadays, the prepend and append import modes use importlib.import_module() instead of __import__().
There was a phrase “which avoids having to use `__import__`”, in which I couldn’t just replace `__import__` by `importlib.import_module` because the latter is used (in insert_missing_modules()) also when using importlib mode. Therefore I removed the part from the sentence.
This makes it possible to correlate pytest stages with external events, and also makes it readable when TestReports are exported externall (for example with pytest-reportlog).
Closes#10710
* [pre-commit.ci] pre-commit autoupdate
updates:
- [github.com/psf/black: 22.12.0 → 23.1.0](https://github.com/psf/black/compare/22.12.0...23.1.0)
- [github.com/PyCQA/autoflake: v2.0.0 → v2.0.1](https://github.com/PyCQA/autoflake/compare/v2.0.0...v2.0.1)
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
* Update .pre-commit-config.yaml
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Bruno Oliveira <nicoddemus@gmail.com>
Instead of `SysCapture`/`FDCapture` inheriting from
`SysCaptureBinary`/`FDCaptureBinary`, have both inherit from a common
`SysCaptureBase`/`FDCaptureBase`. This fixes a Liskov substitution
violation.
Since pytest now requires Python>=3.7, we can use the stdlib attrs
clone, dataclasses, instead of the OG package.
attrs is still somewhat nicer than dataclasses and has some extra
functionality, but for pytest usage there's not really a justification
IMO to impose the extra dependency on users when a standard alternative
exists.
Makes `AbstractContextManager` the shared base class between "raises" and other context managers.
The motivation is for type checkers to narrow `pytest.raises(...) if x else nullcontext()` to a `ContextManager` rather than `object`.
Fixes https://github.com/pytest-dev/pytest/issues/10604 which could intermittently display unexpected behavior between checking if the path exists and requesting creation. This was fairly prevalent when pytest was being invoked in parallel by another test runner (CTest) and trying to create the same parent-folder for multiple XMLs. A modest amount of testing did not reproduce other filesystem race conditions.
This notably does not work around an edge case where the parent path of the XML could be created as a file instead of a folder or link. That vanishingly rare case should cause file creation to fail on the next line, with a fairly obvious exception message.
mypy stubs recently changed warnings.catch_warnings to a Generic, in
order to have proper overloads depending on the parameters passed to it, whihc
triggers this mypy error now when we subclass it:
src/_pytest/recwarn.py:170: error: Missing type parameters for generic type "catch_warnings" [type-arg]
For our porpuses the parameter is not relevant (we always use record=True), so
decided to just ignore the type error.
* Squashed commit of the following:
commit 41d339c46763bbe26123e1e6504b6e32290e33e1
Author: Cheukting <cheukting.ho@gmail.com>
Date: Thu Jun 23 17:01:04 2022 +0800
test in all py versions
commit b3572a5a12672228c3276fc8c8e05980dfb7888a
Author: Cheukting <cheukting.ho@gmail.com>
Date: Thu Jun 23 16:41:06 2022 +0800
add test
commit 7166a2a51e4f99046b028b663c193d8b558c7fd4
Author: Cheukting <cheukting.ho@gmail.com>
Date: Thu Jun 23 16:00:07 2022 +0800
update changelog
commit b958c73d489157f0c0d4e46425083a5e2e2bc851
Author: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Date: Thu Jun 23 07:50:52 2022 +0000
[pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
commit ea7f376c6ca37c40c83df0e4a1cfaaedb34bae91
Author: Cheukting <cheukting.ho@gmail.com>
Date: Thu Jun 23 15:48:21 2022 +0800
Fix MyPy
commit 97469beb1da40257e9a061a5e19548546c9312c4
Author: Cheukting <cheukting.ho@gmail.com>
Date: Thu Jun 23 15:03:48 2022 +0800
fix if ExceptionGroup not exist
commit 84e553642cd69b4d499231d733df91ebfa84c7ad
Author: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Date: Thu Jun 23 03:43:27 2022 +0000
[pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
commit 76bbef449b88bbd74fb5cca3b5293337a624ef03
Author: Cheukting <cheukting.ho@gmail.com>
Date: Thu Jun 23 11:40:41 2022 +0800
adding changelog
commit db82bebc5a4969e2083adcd97bdfd2a63bb17d98
Author: Cheukting <cheukting.ho@gmail.com>
Date: Thu Jun 23 11:33:10 2022 +0800
fall back to native when handeling to exception groups
* Typed ExceptionGroupTypes and changed to BaseExceptionGroup, fixed exceptionchain (excinfo->excinfo_, set reprcrash. Extended tests, though they're wip.
* added exceptiongroup to pre-commit-config, moved away from tuple to directly defining BaseExceptionGroup, added block comment, added match line for inner exception, changked mark.skipif to importorskip to not need top-level import, changed tox.ini a bit - only uncovered should now be py37 without exceptiongroup, due to hypothesis
* added py311-exceptiongroup to github CI, exceptiongroup is now a hard dependency on py<3.11, renamed bad variable names
* added use_coverage to ubuntu-py311
* import BaseExceptionGroup with explicit version check instead of try/catch
* removed from CI, added comments to tox and pre-commit
This should help users with the common issue of patching the wrong place.
Also took the opportunity of using proper links in the monkeypatch introduction.
Related to #10216
Co-authored-by: Ran Benita <ran@unusedvar.com>
The main motivation for this change is to simplify the type shown in
code editors -- `Sequence[str]` is easier to follow than
`Union[list[str], tuple[str, ...]]`.
It also permits using other types if desired. It might lead to problems
if someone uses some oddball sequence type, but hopefully they won't do
that.
New Sphinx added support for overloads and always displays them all with
full type annotations etc. This regresses the API reference for
overloaded functions like `fixture()`, `warns()`, `raises()` and friends
to become impossible to read.
I tried various workarounds but none worked except this one.
New versions of sphinx starting showing `__init__` parameters even when
we don't want them to show because they are private (have `_ispytest`
argument).
The only working solution I found was to switch to
`autodoc_typehints_description_target = "documented"` and explicitly
document parameters for which we want to show the types. It's a little
tedious and repetitive in some simple cases, but overall it results in
nicer API docs.
This started to fail recently with:
```
mypy.....................................................................Failed
- hook id: mypy
- exit code: 1
src/_pytest/assertion/rewrite.py:284: error: Returning Any from function declared to return "TraversableResources" [no-any-return]
Found 1 error in 1 file (checked 219 source files)
```
Not sure why that started failing, but seems like ignoring that error specifically is OK.
Given we are already creating a `PurePath`, just pass the parts directly to it.
This avoids using `os.path.sep`, that although is an official API, seems not to be available in all systems.
Fix#9791
This is especially useful for large repositories (e.g. monorepos) that
use a hierarchical file system organization for nested test paths.
src/*/tests
The implementation uses the standard `glob` module to perform wildcard
expansion in Config.parse().
The related logic that determines whether or not to include 'testpaths'
in the terminal header was previously relying on a weak heuristic: if
Config.args matched 'testpaths', then its value was printed. That
generally worked, but it could also print when the user explicitly used
the same arguments on the command-line as listed in 'testpaths'. Not a
big deal, but it shows that the check was logically incorrect.
Now that 'testpaths' can contain wildcards, it's no longer possible to
perform this simple comparison, so this change also introduces a public
Config.ArgSource enum and Config.args_source attribute that explicitly
names the "source" of the arguments: the command line, the invocation
directory, or the 'testdata' configuration value.
It's better to have the documentation in one place, instead
of having some in the docstring and some additional
information added to the reference documentation in
`reference.rst`.
These pytester utility methods were annotated to only receive `str`
names, but they naturally support os.PathLike values, as well.
This makes writing some pytester calls a little nicer, such as when
creating a directory based on a `.joinpath()` call. We previously needed
to cast that intermediate value to a `str`.
The left/right operands produced when `verbose > 1` should not contain newlines, because they are used to
build the `summary` string. The `assertrepr_compare` function returns a list of lines, and the summary is one of those lines and should not contain newlines itself.
Fix#9742
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Bruno Oliveira <nicoddemus@gmail.com>
The change from `path not in confuctdir.parents` to the `relative_to`
check in 0c98f19231 broke picking up
conftest files when running against an installed package/site-packages.
See the issue for more details.
Fix#9767.
Including the file name is enough to let the user know what the problem is.
The same is not needed for `.ini` files because the error message includes the path to the file by default.
Fix#9730
It is tempting to use `monkeypatch` to replace the other mechanisms in pytester which change global
state: `CwdSnapshot`, `SysModulesSnapshot`, `SysPathsSnapshot`, however those are more delicate
than they look at first glance so leaving those alone for now.
Close#9708
The dummy modules we introduce in `insert_missing_modules` (due to #7856 and #7859)
would cause problems if the dummy modules actually end up replacing modules
which could be imported normally because they are available in `PYTHONPATH`.
Now we attempt to first import the module via normal mechanisms, and only
introduce the dummy modules if the intermediary modules don't actually exist.
Close#9645
In the following
@pytest.mark.parametrize(..., ids=[val])
the ID values are only allowed to be `str`, `float`, `int` or `bool`.
In the following
@pytest.mark.parametrize(..., [val])
@pytest.mark.parametrize(..., [pytest.param(..., id=val])
a different code path is used, which also allows `bytes`, `complex`,
`re.Pattern`, `Enum` and anything with a `__name__`.
In the interest of consistency, use the latter code path for all cases.
Ran Benita
bzoracler
Brian Larsen
Bryan Ricker
Alex
Bruno Oliveira
Kodi Arfer
Ronny Pfannschmidt
Felix Hofstätter
Stefanie Molin
Alessio Izzo
Paul Kehrer
Zac Hatfield-Dodds
Alex Hadley
rdb
Manuel Jacob
Ilya Konstantinov
HTRafal
pre-commit-ci[bot]
Teejay
Jay
Yannick PÉROUX
q0w
Ramsey
s-padmanaban
Kadino
Yusuke Kadowaki
Pierre Sassoulas
Marko Pacak
Daniel Scheffler
Yusuke Kadowaki
Prerak Patel
Daniel Valenzuela
Anthony Sottile
Hugo van Kemenade
Cal Jacobson
Itxaso Aizpurua
Tony Narlock
Gabriel Landau
Thomas Grainger
Simon K
Florian Bruhin
Thierry Moisan
Adam Turner
Gergely Kalmár
John Litborn
Tatiana Ovary
Sviatoslav Sydorenko
holesch
Brandon Chinn
Robert O'Shea
Wolfremium
Nipunn Koorapati
Alice Purcell
zx.qiu
Paul Reece
Chris Wheeler
Zach OBrien
Neil Girdhar
Cheuk Ting Ho
Jon Parise
Pete Baughman
Ezio Melotti
sommersoft
Tim Hoffmann
Babak Keyvani
Peyman Salehi
Shantanu
johnkangw
Tobias Diez
Paul Müller
Kian Eliasi
Samuel Colvin
Florian Best
Rowin
Aaron Berdy
eduardo naufel schettino