merge train: land #10075 and #10077 - #10113
Merged
Merged
Conversation
added 9 commits
September 12, 2026 06:30
(cherry picked from commit 0c5348c)
(cherry picked from commit 8d66646)
(cherry picked from commit 6183863)
Conflicts resolved on the train (#10072/#10074 landed first): - array/concat_reverse.rs, array/sort.rs: kept main's newer structure -- #10072's three-arm linear append in js_array_concat, and the publish_sorted_values / with_sorted_indices refactor -- then applied this change's actual intent to it, replacing every open-coded `header + size_of::<ArrayHeader>()` element base with `crate::array::array_elements_ptr`. That includes four sites in sort.rs that postdate this branch's base and so were never converted here. - array/push_pop.rs: union -- this change's offset-aware accessor plus main's TAG_HOLE -> undefined conversion, which spread needs and this branch predates. - scripts/gc_runtime_root_holders.json: merged structurally, not textually. The PASS1_MARKED re-audit prose is base + main's #10055/#10054 paragraphs + this change's #10060 paragraph, and each `window.sources` pin is taken from the side that actually moved it (gc/census.rs from here, gc/mod.rs from main) so neither is reverted to the base hash. - test-parity/gc_repsel_corpus.txt: both new entries kept.
(cherry picked from commit 1e79ee9)
(cherry picked from commit b27feab)
(cherry picked from commit 176e5ab)
#10077 gives a dense array a queue front offset: `capacity` counts the slots remaining AFTER the front, so logical element zero is no longer the end of `ArrayHeader`. That PR converted every open-coded `header + size_of::<ArrayHeader>()` element base in its own tree — but two places on current `main` postdate its branch point and were never converted: - `json_tape/cached_read.rs` (from #10064, landed in train161) open-codes the base on the hot materialized-array read. A materialized JSON array is an ordinary `GC_TYPE_ARRAY` that user code can `shift()`, so after this train that read would return the wrong element. Routed through `crate::array::array_elements_ptr`. - `array/sort.rs` gained four more open-coded bases with the `publish_sorted_values` / `apply_sorted_indices` refactor. Converted with the conflict resolution in the pick itself; only the rustfmt reflow lands here. Also converts the six raw-handle debt sites #10075's two new test modules introduced. `--no-raise-vs <merge-base>` refuses both a ceiling on a module that was absent at the base (`string/slice_tests.rs`) and a per-module raise (`gc/tests/runtime_roots/string_slice.rs`, 1 -> 3), so the sites are converted rather than recorded: - `gc/tests/runtime_roots/string_slice.rs`: the two post-collection re-reads now run the collection inside `across_mut` / `across_const`, and the kept-slice assertions inside `with_const_ptr`. That also retires the module's one pre-existing site, so its ceiling line is DELETED — the deletion is the receipt — and the recorded total falls 945 -> 944. - `string/slice_tests.rs`: the three argument-position reads become `with_const_ptr`, which is the documented shape for a self-rooting entry point such as `js_string_slice`.
This was referenced Sep 12, 2026
|
Caution Review failedThe pull request is closed. ℹ️ Recent review info⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Advanced Run ID: ⛔ Files ignored due to path filters (1)
📒 Files selected for processing (179)
📝 WalkthroughWalkthroughThis pull request adds dense ChangesArray shift queue and storage
String slicing and suffix cursors
Benchmark and release support
Estimated code review effort: 5 (Critical) | ~90 minutes Change: Bug fix · Severity of issue fixed: Medium Possibly related PRs
✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
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
1 participant
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.
Merge train landing #10075 and #10077 on top of
maindc0d876fe(train162). Both were
CONFLICTINGagainst currentmain.The individual PRs merge as their own branch, so GitHub's close keywords do not
fire — each original is closed with a pointer comment after this lands, and the
issues it names are swept by hand.
0c5348ce5,8d66646ff,6183863dda7c59a837,1e79ee9e4,b27feab6f,176e5ab86The finding: #10077's front offset vs. code that postdates its branch point
#10077 gives a dense array a queue front offset —
capacitycounts theslots remaining AFTER the front, so logical element zero is no longer the end of
ArrayHeader, and every open-codedheader + size_of::<ArrayHeader>()elementbase becomes wrong for an array that has been
shift()ed. The PR converts allof them in its own tree, in both the runtime (
array::storage::array_elements_ptr)and codegen (
LlBlock::array_elements_addr, 36 call sites across 20 files).mainhas moved since that branch point and grew new ones the PR never saw. Iswept the merged tree for every element-base idiom and converted the two that
remained:
json_tape/cached_read.rs— the hot materialized-array read landed byperf(json): read traced materialized arrays without redundant resolution #10064 in train161. A materialized JSON array is an ordinary
GC_TYPE_ARRAYthat user code can
shift(), so without this the read returns the wrongelement. This is a genuine cross-PR interaction, not a style fix.
array/sort.rs— four more bases from thepublish_sorted_values/apply_sorted_indicesrefactor.The only
size_of::<ArrayHeader>()uses left in the tree are allocation-sizevalidations (
param_type_guard.rs) and GC payload arithmetic, not elementbases; codegen has no remaining literal
+8on an array pointer.Conflict resolutions
array/concat_reverse.rs,array/sort.rs— keptmain's newerstructure (fix(runtime): make spread push scale linearly #10072's three-arm linear append in
js_array_concat, and thesort refactor), then applied fix(runtime): make dense Array.shift queue drains linear #10077's intent to it. Taking "theirs" would have
reverted fix(runtime): make spread push scale linearly #10072 back to the
ptr::copy_nonoverlapping+rebuild_array_layout_exactbulk path that madepush(...chunk)quadratic.array/push_pop.rs— union: fix(runtime): make dense Array.shift queue drains linear #10077's offset-aware accessor plusmain'sTAG_HOLE→undefinedconversion, which spread needs and fix(runtime): make dense Array.shift queue drains linear #10077 predates.scripts/gc_runtime_root_holders.json— merged structurally, nottextually. Each
window.sourcespin is taken from the side that actuallymoved it (
gc/census.rsfrom fix(runtime): make dense Array.shift queue drains linear #10077,gc/mod.rsfrommain); a line unionwould have restored
gc/mod.rsto the stale base hash. ThePASS1_MARKEDprose is base +
main's perf(string): non-ASCII charCodeAt and bracket scans become quadratic #10055/perf(string): trim scans and copies the entire interior for constant-size edge whitespace #10054 re-audits + fix(runtime): make dense Array.shift queue drains linear #10077's perf(runtime): Array.shift queue drain reaches 722x Node at 10k elements and times out at 100k #10060 re-audit,so no changed pin loses the justification it is required to carry.
test-parity/gc_repsel_corpus.txt— both new entries kept.string/mod.rs(fix(string): preserve UTF-16 slices and avoid quadratic suffix parsing #10075) — module-declaration list, sorted union.Maintainer fix commits
fix(runtime): make the JSON cached read honour the array front offset—the two conversions above, plus the six raw-handle debt sites fix(string): preserve UTF-16 slices and avoid quadratic suffix parsing #10075's new
test modules introduced.
--no-raise-vs <merge-base>refuses both a ceilingon a module absent at the base (
string/slice_tests.rs) and a per-moduleraise (
gc/tests/runtime_roots/string_slice.rs, 1 → 3), so they areconverted: the post-collection re-reads now run the collection inside
across_mut/across_const, and the argument-position reads becomewith_const_ptr. That also retires the module's one pre-existing site, so itsceiling line is deleted and the recorded total falls 945 → 944
(
--no-raise-vs origin/main: "none raised").chore: bump workspace version to 0.5.1537— neither PR bumped it, whichis the maintainer's job at merge time;
Cargo.lockregenerated with it.Validation
6375 tests, zero failures. Exit codes are captured from each command itself, not
from a pipeline.
Only one lint gate fails, not the usual three. The two API-doc gates are
normally written off as a
CARGO_TARGET_DIRartifact, becausescripts/regen_api_docs.shbuilds withCARGO_TARGET_DIRbut then reads ahardcoded
$ROOT/target/release/perry. Creating that symlink BEFORE the runmakes them genuinely pass with zero drift. Worth knowing: they are not merely
failing, they are destructive — the shell truncates the redirect target
before exec'ing the missing binary, so a run without the symlink leaves
docs/src/api/reference.mdat 0 bytes, a 4229-line deletion sitting in theworktree. The remaining failure, public benchmark evidence freshness, is
genuinely red on
mainsince 2026-07-29 and is scoped out by #9969.The four new gap tests this train adds were run directly against the pinned Node
26.5.1, with the static archives rebuilt first
(
-p perry -p perry-runtime-static -p perry-stdlib-static) and their mtimesconfirmed newer than the last commit:
The two GC fixtures are registered in
test-parity/gc_repsel_corpus.txt, sogc-stress and gc-moving-witnesses will execute them rather than leaving them
dark on disk.
Each suite was also checked for having actually run its subject: 10 tests from
the new
array::shift_queue_tests,string::slice_tests,gc::tests::copying::shift_queueandgc::tests::runtime_roots::string_slicemodules appear in the runtime log.
Validated at head
2aa97f529d45ab7041f5f3fc58f03e9ce45091f3.Summary by CodeRabbit
Bug Fixes
Array.prototype.shift()behavior for dense queues, aliases, garbage collection, growth, and observable edge cases.slice(),substring(), andsubstr()handling at UTF-16 and surrogate boundaries, including lone surrogates and Unicode characters.Performance
Tests