summaryrefslogtreecommitdiff
path: root/tests
Commit message (Collapse)AuthorAgeFilesLines
* docs: more description in balance_xdist_plugin.pyNed Batchelder2021-12-111-3/+19
|
* build: make upgradenedbat/upgrade-with-kitsNed Batchelder2021-11-303-4/+4
|
* test(feat): track test times, and balance xdist workersNed Batchelder2021-11-282-49/+179
| | | | In the end, this seems to only speed things up by about 5%
* test(refactor): avoid full commands to speed testsNed Batchelder2021-11-262-199/+192
| | | | Also btw, I think some of these tests are duplicative.
* test(perf): shave a few seconds off the FailUnder testsNed Batchelder2021-11-261-12/+9
|
* feat: multiple --concurrency values. #1012 #1082nedbat/multi-concurrencyNed Batchelder2021-11-252-17/+96
|
* test(fix): keep SourceIncludeOmitTest's from clobbering each otherNed Batchelder2021-11-241-8/+20
| | | | | | | | Because they cd'd into a folder in the shared source tree, their data files would collide, leading to flaky tests. Also, add a check that the tests collected some data, and add sys.path to the debug=trace output.
* test(fix): skip tests that trip a specific Mac 3.9 eventlet bugNed Batchelder2021-11-241-0/+10
|
* test(build): a better way to pin light-thread packagesNed Batchelder2021-11-241-0/+1
| | | | | Also, clearly indicate when we are skipping tests because the packages aren't available.
* test(refactor): simplify the logic of writing subcover.pthNed Batchelder2021-11-211-4/+2
|
* test(fix): remove the subcover.pth file when the main session endsNed Batchelder2021-11-211-16/+20
| | | | | Leaving the pth file meant that the next igor run would start coverage, which meant igor couldn't delete the C extension file on Windows.
* test(refactor): if two skips apply, put the more significant reason closer ↵Ned Batchelder2021-11-212-2/+2
| | | | to the def
* test(refactor): use pathlib for the pth filesNed Batchelder2021-11-211-11/+9
|
* test(fix): never delete the pth fileNed Batchelder2021-11-211-4/+6
|
* fix(test): make .pth files once per sessionNed Batchelder2021-11-213-70/+55
| | | | | | | | | | | The old way, we made and deleted .pth file around each test. This caused problems because pth files are written to a common location shared by all test workers, so tests would fail because pth files were being deleted as other workers were trying to use them. Now we make the pth file once per session. There's still a chance that a worker will be using a pth file just as another worker is ending and removing it.
* debug: see the output from run_in_venvNed Batchelder2021-11-211-0/+2
|
* refactor(test): move unicode handling tests to individual test filesNed Batchelder2021-11-204-85/+87
| | | | Eight fewer sub-processes
* refactor(test): three more subprocesses gone in test_summary.pyNed Batchelder2021-11-201-22/+10
|
* refactor(test): avoid 8 subprocesses in test_summary.pyNed Batchelder2021-11-201-45/+29
|
* refactor(test): remove subprocesses from test_xml.pyNed Batchelder2021-11-201-15/+23
|
* refactor(test): no more subprocesses in test_html.pyNed Batchelder2021-11-201-9/+13
|
* refactor(test): make_data_file is a helper for making data filesNed Batchelder2021-11-203-19/+27
|
* fix: suffix=False will suppress the suffix even with multiprocessing. #989Ned Batchelder2021-11-181-0/+20
|
* test: properly reset a few globals between every testNed Batchelder2021-11-172-3/+19
|
* refactor: specialize exceptionsNed Batchelder2021-11-148-57/+57
| | | | | | CoverageException is fine as a base class, but not good to use for raising (and catching sometimes). Introduce specialized exceptions that allow third-party tools to integrate better.
* docs: document the exceptionsNed Batchelder2021-11-143-9/+9
| | | | ... and make some of them private.
* test: add a test of `coverage debug premain`Ned Batchelder2021-11-131-0/+15
|
* debug: the `coverage debug` command no longer accepts multiple topicsNed Batchelder2021-11-131-3/+7
|
* feat: 'debug data' now enumerates combinable files alsoNed Batchelder2021-11-131-9/+31
|
* refactor: add_lines only needs an iterable of intsNed Batchelder2021-11-134-31/+31
| | | | | The old code expected a dict with int keys, but iterables are simpler, so get rid of a lot of dict.fromkeys() nonsense.
* test(fix): make the fullcoverage test work on 3.11Ned Batchelder2021-11-121-3/+1
| | | | | It didn't work because the os module is frozen on 3.11, so it gets no measurement. Use -X to disable the frozen modules feature for the test.
* fix: don't overwrite a .gitignore in the html output directory. #1244nedbat/1244-againNed Batchelder2021-11-121-0/+7
|
* test(refactor): convert looping tests to parametrizeNed Batchelder2021-11-113-51/+45
|
* style: convert more string formatting to f-stringsNed Batchelder2021-11-1115-32/+30
|
* refactor: we no longer need to treat 'class' lines speciallyNed Batchelder2021-11-111-3/+3
| | | | | Originally, this tokenizing code was paired with bytecode analysis. But now we use AST instead, so class lines don't need to be handled differently.
* fix: colons in decorators shouldn't stop an exclusionNed Batchelder2021-11-101-1/+21
|
* fix: CPython 3.11 support. #1241Ned Batchelder2021-11-102-0/+4
| | | | | | | | | | | The fix for CTracer is egregious and will need to be updated when there's a supported way to do it. The fullcoverage skip is noted in https://github.com/nedbat/coveragepy/issues/1278 The raise_through_with skip is noted in https://github.com/nedbat/coveragepy/issues/1270
* fix: warn about more source file problemsNed Batchelder2021-11-091-1/+6
|
* refactor(test): convert eight tests to one parametrized testNed Batchelder2021-11-071-27/+4
|
* refactor(test): use the expected name for initializing tests.Ned Batchelder2021-11-079-27/+24
| | | | | | | | | Originally I used setup_test because something went wrong when I used setUp. I wrote https://github.com/pytest-dev/pytest/issues/8424 about it. There they say to use `-p no:nose` to disable nose interpretation. But now I simply went back to setUp, and all seems well? Not sure what changed, but using the expected name is better.
* fix: remapping paths during combining needs to follow relative_files=True. #1147Ned Batchelder2021-11-062-76/+101
|
* fix: give a default value to the name argument for Matchers. #1273Ned Batchelder2021-11-051-6/+6
|
* fix: don't report branches to exclusions as missing. #1271Ned Batchelder2021-11-051-1/+15
|
* refactor(test): make traceback checks a bit flexibleNed Batchelder2021-11-042-7/+11
| | | | | | | | | | | | Python 3.11 made a traceback look like this: Traceback (most recent call last): File "{path}", line 8, in <module> print(sys.argv[1]) ~~~~~~~~^^^ IndexError: list index out of range We needed to not care if that tilde-caret line was present or not.
* refactor(test): don't even import contracts if not using PyContractsNed Batchelder2021-11-041-2/+2
| | | | | | | | On CPython 3.11, PyContracts fails on import, so we need to shut it off more completely. PyContracts is unmaintained these days, so we'll probably get rid of it completely soon.
* test(fix): isolate excepthook tests from each otherNed Batchelder2021-11-022-0/+9
|
* build(test): remove obsolete jQuery gold files #840Ned Batchelder2021-11-026-182/+23
|
* test: force VirtualenvTest onto one worker to save timeNed Batchelder2021-11-021-7/+50
| | | | | | | | | | | VirtualenvTest is slow because it has a session-scoped fixture that takes 10s to run. If all of those tests go to the same worker, we can reuse that fixture. If they go to different workers, we have to spend that time more than once. This is a hack which seems to work, but no guarantees into the future. Also, I don't know why the VirtualenvTests aren't run first on their worker. Time saved: about 10%, from ~50s to ~45s.
* perf(test): run pip just once, shaves 1.2s off the fixtureNed Batchelder2021-10-311-8/+9
| | | | ... though that means it's 8.1s instead of 9.3s, so still pretty slow.
* test: use the coverage-command fixture throughout the venv testsNed Batchelder2021-10-311-4/+4
|
View Lorry Controller Status