GH-107450: fix unexpected OverflowError in testMemoryErrorBigSource #110801
Conversation
|
FYI @lysnikolaou already opened #110768 to fix this :) |
|
Done, and sorry for not noticing the already open issue 😅 |
|
Since I can just rebase or merge main easily, we can still add the improvements. |
|
Ugh, the merge shows the extra changes from main on the Github UI. Next time I'll just rebase and force-push, which should be less confusing 🙃 |
Lib/test/test_exceptions.py
Outdated
| # the padding length needs to be more than INT_MAX, but we can't | ||
| # multiple a sequence by a number that doesn't fit Py_ssize_t, | ||
| # otherwise we will get an OverflowError (see PySequence_Repeat) | ||
| padding = ' ' * ((ctypes.sizeof(ctypes.c_int) + 1) / 8) |
There was a problem hiding this comment.
I don't think this tests the correct thing. The buffer is way too small to overflow, probably a typo?
There was a problem hiding this comment.
Yep, I forgot it returns the number of bytes 😅
Lib/test/test_exceptions.py
Outdated
| @unittest.skipIf(ctypes.sizeof(ctypes.c_ssize_t) > ctypes.sizeof(ctypes.c_int), | ||
| "Py_ssize_t is bigger than INT_MAX, so this is unreachable") |
There was a problem hiding this comment.
This should probably the other way around. If sizeof(ssize_t) <= sizeof(int) the cast is always safe, right?
Lib/test/test_exceptions.py
Outdated
| @unittest.skipIf(ctypes.sizeof(ctypes.c_ssize_t) > ctypes.sizeof(ctypes.c_int), | ||
| "Py_ssize_t is bigger than INT_MAX, so this is unreachable") | ||
| @unittest.skipIf(ctypes.sizeof(ctypes.c_int) >= ctypes.sizeof(ctypes.c_ssize_t), | ||
| "INT_MAX is bigger than Py_ssize_t, so this is unreachable") | ||
| @support.requires_resource('cpu') | ||
| @support.bigmemtest(support._2G, memuse=1.5) | ||
| def testMemoryErrorBigSource(self): |
There was a problem hiding this comment.
This probably needs another merge, since a _size argument has been added. Also, it might be good to request the size of the buffer we want and then use that. Something like:
@support.bigmemtest(2 ** (ctypes.sizeof(ctypes.c_int) * 8), memuse=1)
def testMemoryErrorBigSource(self, size):
padding = ' ' * (size // 8)|
@lysnikolaou is everything addressed, or is there still anything missing? |
|
There is something wrong with the commits in this PR because there are plenty of unrelated changes |
…urce Signed-off-by: Filipe Laíns <[email protected]> add the suggested test markers Signed-off-by: Filipe Laíns <[email protected]> fix incorrect logic Signed-off-by: Filipe Laíns <[email protected]> add _size arg that got lost in the merge Signed-off-by: Filipe Laíns <[email protected]>
|
I think it's just the Github UI showing all the changes from the merge commits, instead of showing the difference with |
|
@FFY00 Thanks for taking time to work on this. Turns out the fix was a bit more complex than expacted, both for the test (which makes the buildbot fail) and for the underlying issue. I've opened #110832 which fixes both issues, so I'd suggest to close this and have you as a co-author there. Are you okay with that? |
1 similar comment
|
@FFY00 Thanks for taking time to work on this. Turns out the fix was a bit more complex than expacted, both for the test (which makes the buildbot fail) and for the underlying issue. I've opened #110832 which fixes both issues, so I'd suggest to close this and have you as a co-author there. Are you okay with that? |
|
Sounds good 👍 |
test_exceptions.ExceptionTests.testMemoryErrorBigSourcefails on certain platforms #110800