[WIP] Fix bug with exception variable reuse in deferred node. #2290
Conversation
This file contains 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
|
@JukkaL Can you think of a way to make the test fail if somehow the function no longer makes it into the deferred_nodes queue? If not I think this fine to merge. |
|
What about having a test option to print out the names of deferred functions at the end of a test case? Deferred type checking has some pretty tricky potential failure modes so good tests are important. Otherwise, some of our tests that are supposedly testing deferred nodes might actually only require a single pass. |
gvanrossum
changed the title
|
Continued discussion in #1748 indicates this fix is wrong! |
gvanrossum
force-pushed
the
deferred-exc
branch
from
971c426
to
1bda6f5
Compare
|
This looks reasonable to me. It's probably not the cleanest fix for this issue but we can live with it. Add a TODO comment about a potentially cleaner approach and this is ready to merge. (Also discussed this offline.) |
Michael0x2a
added a commit
to Michael0x2a/mypy
that referenced
this pull request
Sep 1, 2022
This diff: - Fixes python#8129 - Fixes python#13043 - Fixes python#13167 For more concise repros of these various issues, see the modified test files. But in short, there were two broad categories of errors: 1. Within the deferred pass, we tend to infer 'Any' for the types of different variables instead of the actual type. This interacts badly with our unreachable and disallow-any checks and causes spurious errors. Arguably, the better way of handling this error is to only collect errors during the final pass. I briefly experimented with this approach, but was unable to find a clean, efficient, and non-disruptive way of implementing this. So, I settled for sprinkling in a few more `not self.current_node_deferred` checks. 2. The 'self.msg.disallowed_any_type(...)` call is normally guarded behind a `not self.chk.current_node_deferred` check. However, we were bypassing this check for `except` block assignments because we were deliberately setting that flag to False to work around some bug. For more context, see python#2290. It appears we no longer need this patch anymore. I'm not entirely sure why, but I'm guessing we tightened and fixed the underlying problem with deferred passes some time during the past half-decade.
Michael0x2a
added a commit
to Michael0x2a/mypy
that referenced
this pull request
Sep 2, 2022
This diff: - Fixes python#8129 - Fixes python#13043 - Fixes python#13167 For more concise repros of these various issues, see the modified test files. But in short, there were two broad categories of errors: 1. Within the deferred pass, we tend to infer 'Any' for the types of different variables instead of the actual type. This interacts badly with our unreachable and disallow-any checks and causes spurious errors. Arguably, the better way of handling this error is to only collect errors during the final pass. I briefly experimented with this approach, but was unable to find a clean, efficient, and non-disruptive way of implementing this. So, I settled for sprinkling in a few more `not self.current_node_deferred` checks. 2. The 'self.msg.disallowed_any_type(...)` call is normally guarded behind a `not self.chk.current_node_deferred` check. However, we were bypassing this check for `except` block assignments because we were deliberately setting that flag to False to work around some bug. For more context, see python#2290. It appears we no longer need this patch anymore. I'm not entirely sure why, but I'm guessing we tightened and fixed the underlying problem with deferred passes some time during the past half-decade.
hauntsaninja
pushed a commit
that referenced
this pull request
Sep 2, 2022
…#13575) This diff: - Fixes #8129 - Fixes #13043 - Fixes #13167 For more concise repros of these various issues, see the modified test files. But in short, there were two broad categories of errors: 1. Within the deferred pass, we tend to infer 'Any' for the types of different variables instead of the actual type. This interacts badly with our unreachable and disallow-any checks and causes spurious errors. Arguably, the better way of handling this error is to only collect errors during the final pass. I briefly experimented with this approach, but was unable to find a clean, efficient, and non-disruptive way of implementing this. So, I settled for sprinkling in a few more `not self.current_node_deferred` checks. 2. The `self.msg.disallowed_any_type(...)` call is normally guarded behind a `not self.chk.current_node_deferred` check. However, we were bypassing this check for `except` block assignments because we were deliberately setting that flag to False to work around some bug. For more context, see #2290. It appears we no longer need this patch anymore. I'm not entirely sure why, but I'm guessing we tightened and fixed the underlying problem with deferred passes some time during the past half-decade.
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.
See #1748 (comment)