gh-151722: Defer GC tracking of frozendict.fromkeys() result until fully built - #151967
Closed
tonghuaroot wants to merge 1 commit into
Closed
gh-151722: Defer GC tracking of frozendict.fromkeys() result until fully built#151967tonghuaroot wants to merge 1 commit into
tonghuaroot wants to merge 1 commit into
Conversation
Copy link
Copy Markdown
Contributor
Sorry, something went wrong.
tonghuaroot
requested review from
ericsnowcurrently,
markshannon and
methane
as code owners
June 23, 2026 07:02
…til fully built frozendict.fromkeys() built its result with PyIter_Next() on an already GC-tracked object, so a half-built frozendict was reachable from another thread (using the gc module) and could be observed mutating mid-construction in the free threading build. Untrack the result while it is being filled and re-track it once fully built.
tonghuaroot
force-pushed
the
gh-151722-frozendict-fromkeys-ft-race
branch
from
302e646 to
d45e42d
Compare
June 23, 2026 07:07
Copy link
Copy Markdown
Member
|
Let me take a closer look to see whether this is the right approach. Personally, I’m not a fan of calling _PyObject_GC_UNTRACK. I think it would be better for the object to start out untracked and transition to the tracked state once it is fully initialized. |
Sorry, something went wrong.
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
Reviewers
Assignees
3 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
frozendict.fromkeys() fills its result by iterating the keys and inserting
them one at a time with PyIter_Next(), which runs user code (the iterable's
__next__). In the free-threaded build the result is GC-tracked during this
fill loop, so another thread calling gc.get_objects() can reach and take a
reference to the still half-built frozendict. That breaks the
unique-reference invariant the lock-free insert relies on, tripping
assert(can_modify_dict(mp)) on a debug build (or an unsynchronized write
concurrent with a reader on a release build).
This mirrors gh-151740, which applied the same untrack-during-build /
track-when-complete pattern to frozendict_new() and frozendict_vectorcall().
fromkeys() is the remaining construction path that change did not cover.
This is the separate follow-up for the fromkeys() path; in gh-151740
@corona10 suggested submitting it as its own PR once that one was merged.
Issue: #151722