gh-124855: Fix jit perf (and lejit) #126166
Conversation
Fidget-Spinner
requested review from
ericsnowcurrently and
brandtbucher
as code owners
|
I think we need to add skips to the tests in the Perf trampoline when the jit is activated |
|
I have pushed a bunch of fixes to ensure we run the tests that check for anything perf without the JIT as we do in other places |
Lib/test/test_embed.py
Outdated
| @@ -1768,6 +1768,8 @@ def test_init_set_config(self): | |||
| self.check_all_configs("test_init_set_config", config, | |||
| api=API_ISOLATED) | |||
|
|
|||
| @unittest.skipIf("-D_Py_JIT" in (sysconfig.get_config_var('PY_CORE_CFLAGS') or ''), | |||
There was a problem hiding this comment.
This fails on windows for some reason
There was a problem hiding this comment.
I think with my fix we can actually run the test so I am switching to that
There was a problem hiding this comment.
Hmm I'm abit stumped. No clue why your fix isn't working :(.
There was a problem hiding this comment.
It's because the test is running on isolated mode and therefore it ignores the environment flags
There was a problem hiding this comment.
I pushed a fix that should do the trick
Signed-off-by: Pablo Galindo <[email protected]>
pablogsal
force-pushed
the
fix-jit-perf
branch
from
37e9402
to
8ed27a0
Compare
|
Still failing on Windows... not sure why |
|
Okay, I finally set up my Windows environment and it seems that there is no |
|
Maybe we can land a version of this that just ignores warning lines in stderr and then undo that hack when we have JIT detection on windows. This way we can move on and avoid having to revert |
I put up #126302 as a possible work-around. |
|
Closing it due to merge of #126302. Thank you Pablo and Ken. |
JIT builds are currently failing on main. This supersedes #126163. This wasn't caught because we don't run the JIT tests on
pylifecycle.c