[pull] master from ruby:master#887
Merged
pull[bot] merged 8 commits intoturkdevops:masterfrom Mar 26, 2026
Merged
Conversation
ZJIT: Invalidate ISEQ version on EP escape NoEPEscape PatchPoint side exits use without_locals() to avoid overwriting locals that may have been modified externally (e.g., by eval or instance_eval). This is correct when the PatchPoint fires during a call where EP has actually escaped — the interpreter reads locals from the heap EP, not the stack. However, PatchPoints are globally patched: once EP escapes in any call, the PatchPoint fires on ALL subsequent calls, including ones where EP hasn't escaped. On those calls, the interpreter reads from the stack, where the JIT may have only computed locals in registers (e.g., [] for keyword argument defaults from checkkeyword) without writing them to the frame. The without_locals() side exit doesn't save these register values, so the interpreter sees stale values. This was exposed by the no-profile-send recompilation path, which creates partially-optimized methods where a previously-spilling Send is inlined away, leaving no gen_spill_locals before the PatchPoint. Fix by invalidating the ISEQ version when EP escapes. This resets jit_func so subsequent calls go through the interpreter, which correctly handles all locals from the start. This avoids the stale locals problem without changing the without_locals() optimization.
In mmtk, finalizers are registered as postponed jobs
(rb_postponed_job_preregister) rather than running inline during GC
as with the default GC. Postponed jobs can execute on any Ruby thread
that checks for pending jobs, not just the thread that triggered GC.
This means a popen3 wait_thread can end up running a finalizer for
its own ProcessWrapper, causing wait_thread.value to attempt joining
the current thread, which raises ThreadError ("Target thread must not
be current thread").
Guard against this by skipping the join when the finalizer happens to
run on the wait_thread itself.
Default gem gemspecs list native extension files (e.g. date_core.bundle) without the require path prefix (lib/). In new_format mode, register_default_spec skipped these files because they didn't match the prefix pattern, causing find_default_spec to return nil for them. This led to activation conflicts when a regular gem version of the same gem was also installed: the require fallback path (find_in_unresolved_tree) would find the native extension in the regular gem and try to activate it, conflicting with the already-activated default gem. Fix by also registering non-prefixed files that have a native extension suffix (e.g. .bundle, .so), while still skipping non-requirable files (README, ext/ sources, etc.). ruby/rubygems@023fcae74f Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…fleRuby in TestGemCommandsOwnerCommand ruby/rubygems@924cf41d11 Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
[Bug #21964]
…ordering SimpleCov and test-unit both use at_exit hooks, which Ruby executes in LIFO order. When SimpleCov was loaded after test-unit, its at_exit hook fired first — before tests had run — producing a spurious "Coverage report generated" message from stale results. Moving SimpleCov setup before test-unit ensures its at_exit hook is registered first and therefore runs last, after tests have completed. ruby/rubygems@47e1c347e8 Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Run gem tests concurrently using a thread pool instead of sequentially. Each gem test runs in its own process group with output captured via pipes and printed in original order after all tests complete. Concurrency is controlled by `TEST_BUNDLED_GEMS_NPROCS` env var, defaulting to `[Etc.nprocessors, 8].min`. Per-process `RUBYLIB` is passed via `Process.spawn` env hash to avoid shared `ENV` mutation across threads. Full suite on 10-core machine: 268s → ~107s (2.5x speedup).
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 subscribe to this conversation on GitHub.
Already have an account?
Sign in.
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.
See Commits and Changes for more details.
Created by
pull[bot] (v2.0.0-alpha.4)
Can you help keep this open source service alive? 💖 Please sponsor : )