BUG: fix np.fromiter corruption when reusing a StringDType instance - #32012
Merged
Conversation
added 2 commits
July 14, 2026 10:10
PyArray_FromIter packed values using the caller-supplied ``dtype`` rather
than the descriptor of the array it actually allocated. Those differ when
PyArray_NewFromDescr returns a distinct descriptor -- notably for
StringDType, where reusing an already array-owned instance yields a clone
with its own string arena. Packing then wrote the strings into the
original instance's arena while the array's descriptor referenced a
different one, so on cleanup the data was freed against the wrong
(already-freed) arena, corrupting memory and crashing.
Pack with the array's own descriptor instead, except for subarray dtypes
where the subarray ``dtype`` must still be used.
Yeshwanth-G
changed the title
np.fromiter corruption when reusing a StringDType instance np.fromiter corruption when reusing a StringDType instance
ngoldbaum
reviewed
Jul 16, 2026
Yeshwanth-G
force-pushed
the
fromiter_reuse_segfault
branch
from
July 16, 2026 09:44
534fe8b to
b4174c6
Compare
added 2 commits
July 16, 2026 05:46
Yeshwanth-G
marked this pull request as ready for review
July 16, 2026 09:53
ngoldbaum
approved these changes
Jul 16, 2026
ngoldbaum
left a comment
ngoldbaum
left a comment
Member
There was a problem hiding this comment.
LGTM, just two nits inline.
ngoldbaum
reviewed
Jul 16, 2026
Contributor
Author
|
Thanks for the reviews here @ngoldbaum , @seberg ! |
charris
pushed a commit
that referenced
this pull request
Jul 27, 2026
…ce (#32012) Co-authored-by: Gopu Yeshwanth Reddy <Gopu.Reddy@deshaw.com>
charris
added a commit
that referenced
this pull request
Jul 27, 2026
BUG: fix `np.fromiter` corruption when reusing a `StringDType` instance (#32012)
ngoldbaum
pushed a commit
to ngoldbaum/numpy
that referenced
this pull request
Aug 4, 2026
…ce (numpy#32012) Co-authored-by: Gopu Yeshwanth Reddy <Gopu.Reddy@deshaw.com> (cherry picked from commit adf557d)
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
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.
PR summary
Fixes #32013
np.fromiteris buggy for StringDtype arrays. Specifically, when the dtype instance is already owned by someother array,it creates corrupted structures since it re-uses the input dtype instance to create the output array.
This PR fixes this issue to avoid using the input dtype descriptor to write the output.
Reproducer which this PR fixes:
AI Disclosure
I’ve used Claude Opus to shepherd this along.