summaryrefslogtreecommitdiff
Commit message (Collapse)AuthorAgeFilesLines
* REL: NumPy 1.20.2 release.v1.20.2Charles Harris2021-03-271-1/+1
|
* Merge pull request #18681 from charris/prepare-1.20.2-releaseCharles Harris2021-03-272-31/+75
|\ | | | | REL: Prepare for the NumPy 1.20.2 release.
| * REL: Prepare for the NumPy 1.20.2 release.Charles Harris2021-03-272-31/+75
|/ | | | | - Create the 1.20.2-changelog - Update the 1.20.2-notes
* Merge pull request #18671 from seberg/backport-small-pytest-leaks-fixesCharles Harris2021-03-237-26/+41
|\ | | | | BUG: Fix small issues found with pytest-leaks
| * BUG: Fix small issues found with pytest-leaksSebastian Berg2021-03-237-26/+41
|/ | | | | | | | | | | None of these are particularly worrying as they either usually only leak reference (and not memory) or appear in rare or almost impossible error-paths, or are limited to the tests. Unfortunately, this PR will not apply to 1.20.x, due to small changes in the overrides. Backport of gh-18670 (with two small differences)
* Merge pull request #18661 from charris/backport-18651Charles Harris2021-03-215-1/+6
|\ | | | | BUG: Fix small valgrind-found issues (#18651)
| * BUG: Fix small valgrind-found issues (#18651)Sebastian Berg2021-03-215-1/+6
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | * BUG: Fix small valgrind-found issues This should be backportable. There was at least one that I could not reproduce when running the tests again. And the new random-shuffle tests give false-positives (which is just slightly annoying, considering that we are very close to almost only "longdouble" related false-positives) * BUG: Add missing decref in user-dtype fallback paths The missing decref here only leaks references and can never leak actual memory fortunately. * MAINT,TST: Simplify the "refcount logic" in the dispatch tests again Using SETREF can be nice, but was just overcomplicating thing here...
* | Merge pull request #18638 from charris/backport-18605Charles Harris2021-03-172-2/+23
|\ \ | | | | | | BUG: Fix ma coercion list-of-ma-arrays if they do not cast to bool
| * | Update numpy/ma/tests/test_core.pySebastian Berg2021-03-171-1/+1
| | |
| * | Apply suggestions from code reviewSebastian Berg2021-03-171-1/+2
| | |
| * | BUG: Fix ma coercion list-of-ma-arrays if they do not cast to boolSebastian Berg2021-03-172-2/+22
| |/ | | | | | | | | | | | | | | | | | | | | There was a regression here due to force casting to bool, but if that happens to fail (it does, but should not for strings). The mask would just be dropped. Of course masked arrays are held together by faith here, but its a regression. Closes gh-18551
* | Merge pull request #18636 from BvB93/modules2Charles Harris2021-03-177-2/+49
|\ \ | |/ |/| MAINT: Ensure that re-exported sub-modules are properly annotated
| * API: Move `polynomial.polyutils` to the `PRIVATE_BUT_PRESENT_MODULES` listBas van Beek2021-03-174-4/+1
| | | | | | | | Aforementioned module was accidently marked as public
| * MAINT: Re-export a number of sub-modulesBas van Beek2021-03-176-0/+50
| | | | | | | | | | | | | | Ensures that type checkers will allow the likes of: >>> import numpy as np >>> out = np.lib.stride_tricks.sliding_window_view(...)
| * API: Formally classify `np.lib.stride_tricks` as part of the public APIBas van Beek2021-03-171-1/+1
|/ | | | | | | | | With `as_strided`, and the newly introduced `sliding_window_view` function, there are currently 2 public objects that can: a. Only be imported from a private module b. Are publicly documented to-be imported from aforementioned module Both observations are problematic and in need of rectification. This commit therefore moves `np.lib.stride_tricks` to the `PUBLIC_MODULES` list.
* Merge pull request #18617 from charris/fix-gcc48-testCharles Harris2021-03-151-0/+1
|\ | | | | CI: Update apt package list before Python install
| * CI: Update apt package list before Python installCharles Harris2021-03-151-0/+1
|/ | | | Closes #18613.
* Merge pull request #18611 from charris/backport-18539Charles Harris2021-03-141-2/+2
|\ | | | | BUG: NameError in numpy.distutils.fcompiler.compaq
| * BUG: NameError in numpy.distutils.fcompiler.compaqMichael Lamparski2021-03-141-2/+2
| | | | | | Fix a simple mistake in commit da0497fdf35 which can produce a NameError when installing numpy in MinGW/MSYS2.
* | Merge pull request #18612 from charris/backport-18560Charles Harris2021-03-142-2/+28
|\ \ | | | | | | BUG: Fixed ``where`` keyword for ``np.mean`` & ``np.var`` methods
| * | BUG: Fixed ``where`` keyword for ``np.mean`` & ``np.var`` methods (gh-18560)Mateusz Sokół2021-03-142-2/+28
| |/ | | | | | | | | | | | | | | | | | | | | * Fixed keyword bug * Added test case * Reverted to original notation * Added tests for var and std Closes gh-18552
* | MAINT: Add annotations for `dtype.__getitem__`, `__mul__` and `names` (#18599)Bas van Beek2021-03-133-0/+55
| |
* | Merge pull request #18546 from ahaldane/backport_18530Charles Harris2021-03-052-1/+14
|\ \ | | | | | | BUG: incorrect error fallthrough in nditer
| * | BUG: incorrect error fallthrough in nditerAllan Haldane2021-03-042-1/+14
| |/ | | | | | | Fixup to gh-18450
* | Merge pull request #18559 from charris/backport-ci-fixesCharles Harris2021-03-052-4/+3
|\ \ | |/ |/| CI: Backport CI fixes from main.
| * CI: Use Ubuntu 18.04 to run "full" test.Charles Harris2021-03-051-1/+1
| | | | | | | | | | | | NumPy does not build using the `--coverage` flag on Ubuntu 20.04, the problem seems to be gcc 9.3.0-17. Work around that by running on Ubuntu 18.04 instead.
| * CI: Pin docker image.Charles Harris2021-03-051-3/+2
|/ | | | | | | | The azure-pipeline test Linux_Python_38_32bit_full_with_asserts has been failing after the release of the latest manylinux2010. Pin the docker image to an earlier version to fix this. Closes #18553.
* Merge pull request #18488 from charris/backport-18479Charles Harris2021-02-241-1/+6
|\ | | | | BUG: check if PyArray_malloc succeeded
| * Goto done when an error condition is reachedChristoph Gohlke2021-02-241-0/+1
| |
| * BUG: check if PyArray_malloc succeededChristoph Gohlke2021-02-241-1/+5
| |
* | Merge pull request #18487 from charris/backport-18478Charles Harris2021-02-241-1/+1
|\ \ | | | | | | BUG: verify pointer against NULL before using it
| * | MAINT: verify pointer against NULL before using itChristoph Gohlke2021-02-241-1/+1
| |/
* | Merge pull request #18486 from charris/backport-18477Charles Harris2021-02-241-1/+1
|\ \ | | | | | | BUG: remove nonsensical comparison of pointer < 0
| * | MAINT: remove nonsensical comparison of pointer < 0Christoph Gohlke2021-02-241-1/+1
| |/
* | Merge pull request #18485 from charris/backport-18476Charles Harris2021-02-241-2/+2
|\ \ | |/ |/| BUG: Remove suspicious type casting
| * MAINT: cast Py_ssize_t to intChristoph Gohlke2021-02-241-1/+1
| |
| * MAINT: Remove suspicious type castingChristoph Gohlke2021-02-241-1/+1
|/
* Merge pull request #18469 from charris/backport-18450Charles Harris2021-02-224-8/+35
|\ | | | | BUG: Segfault in nditer buffer dealloc for Object arrays
| * BUG: Segfault in nditer buffer dealloc for Object arraysAllan Haldane2021-02-224-8/+35
|/
* Merge pull request #18462 from charris/backport-15392Charles Harris2021-02-213-81/+153
|\ | | | | BUG: Remove temporary change of descr/flags in VOID functions
| * TST: Add test for nonzero and copyswapn (through advanced indexing)Sebastian Berg2021-02-212-0/+45
| |
| * BUG: Remove temporary change of descr/flags in VOID functionsSebastian Berg2021-02-211-81/+108
| | | | | | | | | | | | | | | | | | | | | | This is done using a hack around stack allocated arrays, which requires some dark magic with respect to the setting the base object (we skip normal arrays and INCREF/DECREF during base setting, which _does_ work for such static arrays, but only if we give them a (not quite) fake refcount of 1 and set the base and their type correctly to PyArrayType. Closes gh-15387
* | Merge pull request #18461 from charris/backport-18420Charles Harris2021-02-211-0/+7
|\ \ | | | | | | BUG: Fix tiny memory leaks when ``like=`` overrides are used
| * | BUG: Fix tiny memory leaks when `like=` overrides are usedSebastian Berg2021-02-211-0/+7
| |/ | | | | | | | | | | | | I thought I had fixed these leaks, but it appears I missed some. We probably should backport this to 1.20.x (its simple), but the leaks are also pretty harmless unless someone uses `like=` hundrets of thousands of times in a running program (and its a new fetaure).
* | Merge pull request #18460 from charris/backport-18436Charles Harris2021-02-211-0/+1
|\ \ | | | | | | BUG: Fix refcount leak in f2py ``complex_double_from_pyobj``.
| * | BUG: Fix refcount leak in f2py `complex_double_from_pyobj`Sebastian Berg2021-02-211-0/+1
| |/
* | Merge pull request #18459 from charris/backport-18455Charles Harris2021-02-211-4/+4
|\ \ | |/ |/| BUG: ``diagflat`` could overflow on windows or 32-bit platforms
| * BUG: Fixed an issue where `diagflat` could overflow on windows and 32-bit ↵Bas van Beek2021-02-211-4/+4
|/ | | | platforms
* Merge pull request #18382 from charris/update-1.20.x-f2pyCharles Harris2021-02-0912-101/+183
|\ | | | | MAINT: Update f2py from master.
| * MAINT: Update f2py from master.Charles Harris2021-02-0912-101/+183
|/ | | | | This will make it easier to backport the f2py fixes currently going into master.
View Lorry Controller Status