Fix crash involving explicit reexport, import cycle, wildcard #11632
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
added 2 commits
November 28, 2021 02:27
JukkaL
approved these changes
Dec 3, 2021
ilevkivskyi
pushed a commit
that referenced
this pull request
Dec 3, 2021
Fixes #8481 Fixes #9941 Fixes #11025 Fixes #11038 This is the unresolved crash that's been reported the most times, e.g., it's as easy to repro as `mypy --no-implicit-reexport -c 'import pytorch_lightning'` (or even `import torch`, with the right PyTorch version). I know of multiple popular Python packages that have made changes just to work around this crash. I would love to see this released, potentially along with #11630. Thanks to @rraval for making a minimal repro! The fix ended up being a one-liner, but it took me a bit to track down :-) Since things worked with implicit reexport, I differentially patched in explicit reexport logic to narrow things down. This let me observe that if I hardcoded pkg.a.B to get reexported, we hit this branch, which clobbers the PlaceholderNode for pkg.c.B, which fixes things: https://github.com/python/mypy/blob/f79e7afec2c863c34d7a9b41ebb732dc26128fff/mypy/semanal.py#L2028 Which is a little weird — we shouldn't have a PlaceholderNode for pkg.c.B at all. But with a breakpoint in that branch, it was easy to notice that with `--no-implicit-reexport` pkg.a.B was first created with `module_public=True` (resulting in creation of a PlaceholderNode in pkg.c.B) and only on a later pass acquired `module_public=False`. So tracking down where pkg.a.B symbol was first created with `module_public=True` led me to this "obvious" bug.
tushar-deepsource
pushed a commit
to DeepSourceCorp/mypy
that referenced
this pull request
Jan 20, 2022
…#11632) Fixes python#8481 Fixes python#9941 Fixes python#11025 Fixes python#11038 This is the unresolved crash that's been reported the most times, e.g., it's as easy to repro as `mypy --no-implicit-reexport -c 'import pytorch_lightning'` (or even `import torch`, with the right PyTorch version). I know of multiple popular Python packages that have made changes just to work around this crash. I would love to see this released, potentially along with python#11630. Thanks to @rraval for making a minimal repro! The fix ended up being a one-liner, but it took me a bit to track down :-) Since things worked with implicit reexport, I differentially patched in explicit reexport logic to narrow things down. This let me observe that if I hardcoded pkg.a.B to get reexported, we hit this branch, which clobbers the PlaceholderNode for pkg.c.B, which fixes things: https://github.com/python/mypy/blob/f79e7afec2c863c34d7a9b41ebb732dc26128fff/mypy/semanal.py#L2028 Which is a little weird — we shouldn't have a PlaceholderNode for pkg.c.B at all. But with a breakpoint in that branch, it was easy to notice that with `--no-implicit-reexport` pkg.a.B was first created with `module_public=True` (resulting in creation of a PlaceholderNode in pkg.c.B) and only on a later pass acquired `module_public=False`. So tracking down where pkg.a.B symbol was first created with `module_public=True` led me to this "obvious" bug.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Fixes #8481
Fixes #9941
Fixes #11025
Fixes #11038
This is the unresolved crash that's been reported the most times, e.g., it's as easy to repro as
mypy --no-implicit-reexport -c 'import pytorch_lightning'(or evenimport torch, with the right PyTorch version). I know of multiple popular Python packages that have made changes just to work around this crash. I would love to see this released, potentially along with #11630.Thanks to @rraval for making a minimal repro!
The fix ended up being a one-liner, but it took me a bit to track down :-)
Since things worked with implicit reexport, I differentially patched in explicit reexport logic to narrow things down. This let me observe that if I hardcoded pkg.a.B to get reexported, we hit this branch, which clobbers the PlaceholderNode for pkg.c.B, which fixes things:
mypy/mypy/semanal.py
Line 2028 in f79e7af
Which is a little weird — we shouldn't have a PlaceholderNode for pkg.c.B at all. But with a breakpoint in that branch, it was easy to notice that with
--no-implicit-reexportpkg.a.B was first created withmodule_public=True(resulting in creation of a PlaceholderNode in pkg.c.B) and only on a later pass acquiredmodule_public=False. So tracking down where pkg.a.B symbol was first created withmodule_public=Trueled me to this "obvious" bug.