add --disable-plugin-autoload #13253
Conversation
|
I don't have answers for those questions, sorry, but I'm happy to say that the PR is looking good overall, and I'm looking forward to merging it once they're resolved and CI is green 🙂 |
psf-chronographer
bot
added
the
bot:chronographer:provided
|
I can't repro the pypy fail locally, but I'm not sure if spec matters, so ?? I've updated comments to something that almost makes sense to merge with, but ... it's a mess. |
There was a problem hiding this comment.
This looks good to me, but I'd appreciate a second opinion before merging. Maybe @nicoddemus?
Will do, likely tomorrow! 👍 |
| if mode == "plain": | ||
| expected = "E AssertionError" | ||
| elif mode == "rewrite": | ||
| expected = "*assert 10 == 30*" | ||
| else: | ||
| assert 0 | ||
| result.stdout.fnmatch_lines([expected]) | ||
|
|
||
| if not disable_plugin_autoload or explicit_specify: |
There was a problem hiding this comment.
With all the additional logic in this test, I wonder if it would be better to have a fixture for the hampkg and split this into two distinct tests at least (autoload enabled/disabled)?
Co-authored-by: Florian Bruhin <[email protected]>
testing/test_assertion.py
Outdated
| # FIXME: if it's already loaded then you get a ValueError: "Plugin already | ||
| # registered under a different name." | ||
| # vaguely related to https://github.com/pytest-dev/pytest/issues/5661 | ||
| name = "spamplugin" if disable_plugin_autoload else "spam" |
There was a problem hiding this comment.
The difference here is that when loading via entry point, the plugin can choose its own name (name = 'spam'), but when loading via other methods the name of the module is used (spamplugin).
I have simplified things here by always using spamplugin as the name of the plugin (b157cef).
fixes #8969
I'm pretty sure the implementation is solid, but I have a bunch of question marks in comments that should be resolved before merging. I can probably resolve them on my own by digging deep enough into the innards, but wouldn't mind explanations if somebody is more familiar with that territory.
I'm not entirely sure that
test_installed_plugin_rewriteshould be parametrized to hell, as opposed to creating a dedicated test.I also found a random unrelated typo in
test_installed_plugin_rewritewithcheck_first->check_first2, so fixed that + improved thefnmatch_lines.changelogfolder, with a name like<ISSUE NUMBER>.<TYPE>.rst. See changelog/README.rst for details.