fix(stdlib): build with --no-default-features again, unbreaking the auto-optimize relink (#7764) - #7772
Open
proggeramlug wants to merge 1 commit into
Open
fix(stdlib): build with --no-default-features again, unbreaking the auto-optimize relink (#7764)#7772proggeramlug wants to merge 1 commit into
proggeramlug wants to merge 1 commit into
Conversation
📝 WalkthroughWalkthrough
Changesstdlib feature compatibility
Estimated code review effort: 2 (Simple) | ~10 minutes Possibly related PRs
Suggested labels: 🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✨ 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 |
…uto-optimize relink (#7764)
proggeramlug
force-pushed
the
fix/7764-stdlib-no-default-features
branch
from
August 10, 2026 13:29
85c5616 to
60ed63a
Compare
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 4
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@crates/perry-stdlib/src/worker_threads.rs`:
- Around line 973-974: The immediate-termination path around
js_promise_new_for_native_resolution must balance the Promise pin when
resolved_now is true. Route creation and settlement through the async_shim in
both branches, or invoke unpin_promise_after_native_resolution() before the
direct perry_runtime::js_promise_resolve() only under async-runtime, while
preserving normal pending-worker processing.
In `@crates/perry-stdlib/src/worker_threads/async_shim.rs`:
- Around line 44-50: The async-runtime Promise lifecycle is unbalanced between
deferred and immediate settlement paths. Update
js_promise_new_for_native_resolution in
crates/perry-stdlib/src/worker_threads/async_shim.rs:44-50 and the immediate
js_promise_resolve path in crates/perry-stdlib/src/worker_threads.rs:973-974 so
the Promise remains rooted until queued worker resolutions finish, or is
explicitly unpinned before immediate settlement, matching the bridge contract.
- Around line 52-73: Update queue_deferred_resolution and the corresponding
async bridge so converter is never executed on the spawned
wait_for_direct_message_ack thread. Dispatch the conversion to the
main-thread/runtime context, then call queue_promise_resolution with the
converted value there; preserve the existing success/rejection behavior and
call-site compatibility.
- Around line 39-43: Update the no-async-runtime inline ensure_pump_registered()
path so it registers the stdlib worker pump and active-handle hooks instead of
remaining a no-op. Keep Tokio-specific runtime setup gated by async-runtime, and
preserve the unconditional callers from js_worker_threads_worker_new() and
js_worker_threads_message_channel_new() so queued worker events, parentPort
messages, and termination promises continue draining.
🪄 Autofix
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro Plus
Run ID: b939eadb-f308-4f5e-a8de-eb2cebb22ded
📒 Files selected for processing (8)
changelog.d/7772-stdlib-no-default-features.mdcrates/perry-stdlib/src/common/dispatch/init.rscrates/perry-stdlib/src/worker_threads.rscrates/perry-stdlib/src/worker_threads/async_shim.rscrates/perry-stdlib/src/worker_threads/broadcast_channel.rscrates/perry-stdlib/src/worker_threads/direct_message.rscrates/perry-stdlib/src/worker_threads/message_port.rscrates/perry-stdlib/src/worker_threads/worker_pump.rs
Comment on lines
+973
to
+974
| let promise = | ||
| unsafe { crate::worker_threads::async_shim::js_promise_new_for_native_resolution() }; |
There was a problem hiding this comment.
🩺 Stability & Availability | 🟠 Major | ⚡ Quick win
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
set -euo pipefail
echo "Files of interest:"
fd -a 'worker_threads\.rs|async_bridge\.rs|async_shim\.rs' . | sed 's#^\./##'
echo
echo "Inspect worker_threads around lines 950-1015:"
if [ -f crates/perry-stdlib/src/worker_threads.rs ]; then
sed -n '940,1025p' crates/perry-stdlib/src/worker_threads.rs | cat -n
fi
echo
echo "Search for async_shim symbols:"
rg -n "queue_promise_resolution|js_promise_resolve|js_promise_new_for_native_resolution|promise_unpin|native_resolution|pin|resolve" crates/perry-stdlib/src crates -g '*.rs' | head -n 200Repository: PerryTS/perry
Length of output: 26635
🏁 Script executed:
#!/bin/bash
set -euo pipefail
echo "async_shim outline:"
ast-grep outline crates/perry-stdlib/src/worker_threads/async_shim.rs --view expanded | sed -n '1,220p' || true
echo
echo "async_shim relevant content:"
sed -n '1,260p' crates/perry-stdlib/src/worker_threads/async_shim.rs | cat -n
echo
echo "async_bridge outline:"
ast-grep outline crates/perry-stdlib/src/common/async_bridge.rs --view expanded | sed -n '1,240p' || true
echo
echo "async_bridge relevant content:"
sed -n '1,260p' crates/perry-stdlib/src/common/async_bridge.rs | cat -n
echo
echo "worker_threads module/header relevant content:"
sed -n '1,80p' crates/perry-stdlib/src/worker_threads.rs | cat -n
sed -n '680,790p' crates/perry-stdlib/src/worker_threads.rs | cat -nRepository: PerryTS/perry
Length of output: 28996
🏁 Script executed:
#!/bin/bash
set -euo pipefail
echo "Inspect async_bridge resolution/settle paths:"
sed -n '350,430p' crates/perry-stdlib/src/common/async_bridge.rs | cat -n
echo
echo "Inspect process_pending and pending drain:"
sed -n '440,530p' crates/perry-stdlib/src/common/async_bridge.rs | cat -n
echo
echo "Search unpin_promise_after_native_resolution:"
rg -n "unpin_promise_after_native_resolution|js_stdlib_process_pending" crates/perry-stdlib/src crates -g '*.rs'Repository: PerryTS/perry
Length of output: 17175
Balance the native-resolution promise on immediate termination.
When async-runtime is enabled, js_promise_new_for_native_resolution() pins the Promise. The resolved_now branch calls perry_runtime::js_promise_resolve() directly, so unpin_promise_after_native_resolution() is never run through js_stdlib_process_pending(). Repeated termination of exited workers leaves pinned Promises that cannot be collected. Route native Promise creation and settlement through async_shim in both branches, or add a branch that directly calls unpin_promise_after_native_resolution() before resolve only for the async-runtime path.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@crates/perry-stdlib/src/worker_threads.rs` around lines 973 - 974, The
immediate-termination path around js_promise_new_for_native_resolution must
balance the Promise pin when resolved_now is true. Route creation and settlement
through the async_shim in both branches, or invoke
unpin_promise_after_native_resolution() before the direct
perry_runtime::js_promise_resolve() only under async-runtime, while preserving
normal pending-worker processing.
Source: MCP tools
Comment on lines
+39
to
+43
| #[cfg(not(feature = "async-runtime"))] | ||
| mod inline { | ||
| /// No bridge means no pump to register. | ||
| pub(crate) fn ensure_pump_registered() {} | ||
|
|
There was a problem hiding this comment.
🩺 Stability & Availability | 🟠 Major | ⚡ Quick win
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
set -euo pipefail
rg -n -C 8 \
'ensure_pump_registered|js_register_stdlib_pump|js_register_stdlib_has_active|js_worker_threads_(process_pending|has_pending|channels_process_pending|channels_has_pending)' \
crates/perry-stdlib crates/perry-runtimeRepository: PerryTS/perry
Length of output: 50369
🏁 Script executed:
#!/bin/bash
set -euo pipefail
printf '--- worker async_shim feature gates ---\n'
sed -n '1,140p' crates/perry-stdlib/src/worker_threads/async_shim.rs
printf '\n--- worker pump outline and relevant sections ---\n'
ast-grep outline crates/perry-stdlib/src/worker_threads/worker_pump.rs --view expanded || true
rg -n -C 8 'js_worker_threads_(process_pending|has_pending)|start_worker|start_stdin_reader|worker_threads_process_pending|worker_threads_has_pending|ensure_pump_registered|queue_promise'| crates/perry-stdlib/src/worker_threads src/worker_threads crates/perry-runtime/src/common crates/perry-runtime/src/event_pump 2>/dev/null || true
printf '\n--- async_bridge registrations and active/pending methods ---\n'
sed -n '430,475p' crates/perry-stdlib/src/common/async_bridge.rs
rg -n -C 5 'js_register_wait_driver|ensure_gc_scanner_registered|js_stdlib_init_dispatch|ensure_pump_registered|js_run_stdlib_pump|js_stdlib_process_pending|js_stdlib_has_active_handles' crates/perry-runtime crates -g '*.rs' | head -240
printf '\n--- Cargo features for perry-stdlib/runtime ---\n'
rg -n 'async-runtime|perry-stdlib|perry-runtime' Cargo.toml crates -g 'Cargo.toml' | head -160Repository: PerryTS/perry
Length of output: 242
🏁 Script executed:
#!/bin/bash
set -euo pipefail
printf '%s\n' '--- worker async_shim feature gates ---'
sed -n '1,140p' crates/perry-stdlib/src/worker_threads/async_shim.rs
printf '%s\n' ''
printf '%s\n' '--- worker pump outline and relevant sections ---'
ast-grep outline crates/perry-stdlib/src/worker_threads/worker_pump.rs --view expanded || true
rg -n -C 8 'js_worker_threads_(process_pending|has_pending)|start_worker|start_stdin_reader|worker_threads_process_pending|worker_threads_has_pending|ensure_pump_registered|queue_promise' \
crates/perry-stdlib/src/worker_threads crates/perry-runtime/src/common crates/perry-runtime/src/event_pump 2>/dev/null || true
printf '%s\n' ''
printf '%s/n' '--- async_bridge registrations and active/pending methods ---'
sed -n '430,475p' crates/perry-stdlib/src/common/async_bridge.rs
rg -n -C 5 'js_register_wait_driver|ensure_gc_scanner_registered|js_stdlib_init_dispatch|ensure_pump_registered|js_run_stdlib_pump|js_stdlib_process_pending|js_stdlib_has_active_handles' \
crates/perry-runtime crates -g '*.rs' | head -240
printf '%s\n' ''
printf '%$s\n' '--- Cargo features for perry-stdlib/runtime ---'
rg -n 'async-runtime|perry-stdlib|perry-runtime' Cargo.toml crates -g 'Cargo.toml' | head -160Repository: PerryTS/perry
Length of output: 46426
Provide a no-default worker pump registration path.
js_worker_threads_worker_new() and js_worker_threads_message_channel_new() call worker_threads::async_shim::ensure_pump_registered() unconditionally. Without async-runtime, the inline shim is a no-op, so worker events/parentPort messages and termination promises can remain queued, and the event loop can exit before they drain. Keep the Tokio runtime setup behind async-runtime, but still register the stdlib pump and active-handle hooks in the no-default path.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@crates/perry-stdlib/src/worker_threads/async_shim.rs` around lines 39 - 43,
Update the no-async-runtime inline ensure_pump_registered() path so it registers
the stdlib worker pump and active-handle hooks instead of remaining a no-op.
Keep Tokio-specific runtime setup gated by async-runtime, and preserve the
unconditional callers from js_worker_threads_worker_new() and
js_worker_threads_message_channel_new() so queued worker events, parentPort
messages, and termination promises continue draining.
Source: MCP tools
Comment on lines
+44
to
+50
| /// # Safety | ||
| /// Mirrors `async_bridge::js_promise_new_for_native_resolution`. | ||
| /// | ||
| /// No pinning: pinning guards the deferral window, and there is none here. | ||
| pub(crate) unsafe fn js_promise_new_for_native_resolution() -> *mut perry_runtime::Promise { | ||
| perry_runtime::js_promise_new() | ||
| } |
There was a problem hiding this comment.
🩺 Stability & Availability | 🟠 Major | 🏗️ Heavy lift
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
set -euo pipefail
echo "Files:"
git ls-files | rg 'crates/perry-stdlib/src/(worker_threads(_test)?|common/async_bridge\.rs)' || true
echo
echo "async_shim outline:"
ast-grep outline crates/perry-stdlib/src/worker_threads/async_shim.rs --view expanded || true
echo
echo "worker_threads relevant sections:"
wc -l crates/perry-stdlib/src/worker_threads.rs crates/perry-stdlib/src/worker_threads/async_shim.rs
sed -n '1,90p' crates/perry-stdlib/src/worker_threads/async_shim.rs
sed -n '930,990p' crates/perry-stdlib/src/worker_threads.rs
echo
echo "Search promise pins/unpins and js_promise_new:"
rg -n "js_promise_new|js_promise_(pin|unpin)|Promise|native_resolution|resolve" crates/perry-stdlib/src crates/perry-rt crates/perry-runtime crates/perry-rt-bindings 2>/dev/null | head -n 200Repository: PerryTS/perry
Length of output: 28993
🏁 Script executed:
#!/bin/bash
set -euo pipefail
echo "Locate async_bridge docs and perry_runtime Promise APIs:"
rg -n "js_promise_new|js_promise_new_for_native_resolution|pinning|deferral|queued settlement|unpin" crates/perry-stdlib/src crates 2>/dev/null | head -n 300
fd -a 'async_bridge\.rs' .
fd -a -i 'promise|worker_threads' . | head -n 200Repository: PerryTS/perry
Length of output: 40429
🏁 Script executed:
#!/bin/bash
set -euo pipefail
echo "async_bridge full:"
cat -n crates/perry-stdlib/src/common/async_bridge.rs | sed -n '1,110p'
cat -n crates/perry-stdlib/src/common/async_bridge.rs | sed -n '470,555p'
echo
echo "worker_threads full context around promised termination:"
cat -n crates/perry-stdlib/src/worker_threads.rs | sed -n '960,1025p'
echo
echo "direct_message call sites:"
cat -n crates/perry-stdlib/src/worker_threads/direct_message.rs | sed -n '1,140p'
echo
echo "promises tests related to 859 native pin:"
cat -n test-files/test_issue_859_native_promise_pin.ts | sed -n '1,220p'Repository: PerryTS/perry
Length of output: 22648
Keep the bridge Promise contract balanced for async-runtime too.
With async-runtime, js_promise_new_for_native_resolution() now pins the Promise, but the inline immediate-settlement path still resolves it without unpinning. Keep deferral/non-deferral behavior in sync: preserve a GC root until all queued worker resolutions complete, or unpin before the immediate js_promise_resolve() path in the async-runtime fork as well.
📍 Affects 2 files
crates/perry-stdlib/src/worker_threads/async_shim.rs#L44-L50(this comment)crates/perry-stdlib/src/worker_threads.rs#L973-L974
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@crates/perry-stdlib/src/worker_threads/async_shim.rs` around lines 44 - 50,
The async-runtime Promise lifecycle is unbalanced between deferred and immediate
settlement paths. Update js_promise_new_for_native_resolution in
crates/perry-stdlib/src/worker_threads/async_shim.rs:44-50 and the immediate
js_promise_resolve path in crates/perry-stdlib/src/worker_threads.rs:973-974 so
the Promise remains rooted until queued worker resolutions finish, or is
explicitly unpinned before immediate settlement, matching the bridge contract.
Source: MCP tools
Comment on lines
+52
to
+73
| /// Settle now rather than queueing for a pump that does not exist. | ||
| pub(crate) fn queue_promise_resolution(promise_ptr: usize, is_success: bool, result_bits: u64) { | ||
| if promise_ptr == 0 { | ||
| return; | ||
| } | ||
| let promise = promise_ptr as *mut perry_runtime::Promise; | ||
| let value = f64::from_bits(result_bits); | ||
| if is_success { | ||
| perry_runtime::js_promise_resolve(promise, value); | ||
| } else { | ||
| perry_runtime::js_promise_reject(promise, value); | ||
| } | ||
| } | ||
|
|
||
| /// As above, running the converter inline. The `Send + 'static` bound is | ||
| /// kept so the two configurations accept the same call sites. | ||
| pub(crate) fn queue_deferred_resolution<F>(promise_ptr: usize, is_success: bool, converter: F) | ||
| where | ||
| F: FnOnce() -> u64 + Send + 'static, | ||
| { | ||
| queue_promise_resolution(promise_ptr, is_success, converter()); | ||
| } |
There was a problem hiding this comment.
🩺 Stability & Availability | 🔴 Critical | 🏗️ Heavy lift
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
set -euo pipefail
rg -n -C 10 \
'queue_(deferred_)?resolution|std::thread::spawn|worker_messaging_error_value|js_promise_(resolve|reject)|js_string_from_bytes|js_error_new' \
crates/perry-stdlib/src/worker_threads
rg -n -C 8 \
'thread-local arenas|main thread|converter' \
crates/perry-stdlib/src/common/async_bridge.rsRepository: PerryTS/perry
Length of output: 36754
🏁 Script executed:
#!/bin/bash
set -euo pipefail
printf '--- Rust files under perry-runtime/thread and promise ---\n'
git ls-files -- crates/perry-runtime | rg 'thread|promise|value|gc' | sed -n '1,200p'
printf '\n--- Runtime imports in async_bridge.rs ---\n'
rg -n 'js_(resolve|reject|promise_)|thread-local|arena|RuntimeRootVisitor|RuntimeHandleScope|main_thread' crates/perry-stdlib/src/common/async_bridge.rs crates/perry-runtime | sed -n '1,240p'Repository: PerryTS/perry
Length of output: 242
🏁 Script executed:
#!/bin/bash
set -euo pipefail
printf '%s\n' '--- Runtime files likely to contain thread arena/runtime context ---'
git ls-files -- crates/perry-runtime | rg 'thread|promise|value|gc|main' | sed -n '1,200p' || true
printf '%s\n' ''
printf '%s\n' '--- Async bridge import/context ---'
sed -n '1,80p' crates/perry-stdlib/src/common/async_bridge.rs
printf '%s\n' ''
printf '%s\n' '--- Direct message rejection path with spawned worker worker ack ---'
sed -n '130,180p' crates/perry-stdlib/src/worker_threads/direct_message.rsRepository: PerryTS/perry
Length of output: 13680
🏁 Script executed:
#!/bin/bash
set -euo pipefail
printf '%s\n' '--- thread.rs arena definitions/usages ---'
sed -n '1,240p' crates/perry-runtime/src/thread.rs
printf '%s\n' ''
printf '%s\n' '--- promise API/resolve/reject definitions ---'
rg -n -C 8 'fn js_promise_(new|resolve|reject|resol|rejec)|js_promise_new_for_native_resolution|RuntimeHandleScope|RuntimeRootVisitor|ThreadLocal|thread_local!' crates/perry-runtime/src/promise crates/perry-runtime/src/gc/roots crates/perry-runtime/src/gc/mod.rs | sed -n '1,260p'
printf '%s\n' ''
printf '%s\n' '--- inline shim resolution functions ---'
sed -n '4,85p' crates/perry-stdlib/src/worker_threads/async_shim.rsRepository: PerryTS/perry
Length of output: 39964
🏁 Script executed:
#!/bin/bash
set -euo pipefail
printf '%s\n' '--- locate RuntimeHandleScope definitions ---'
rg -n -C 8 'struct RuntimeHandleScope|impl RuntimeHandleScope|enum RuntimeHandleScope|RuntimeHandleScope' crates/perry-runtime/src/gc/roots crates/perry-runtime/src/gc/mod.rs crates/perry-runtime/src/gc | sed -n '1,220p'
printf '%s\n' ''
printf '%s\n' '--- locate thread-local arena initializers and allocation symbols ---'
rg -n -C 5 'thread_local|THREAD_LOCAL|PERRY_THREAD_LOCAL|gc_init|GcInit|Arena|alloc|malloc|thread_spawn|worker_thread|spawn' crates/perry-runtime/src/gc/mod.rs crates/perry-runtime/src/gc/roots.rs crates/perry-runtime/src/gc/roots crates/perry-runtime/src/thread.rs crates/perry-runtime/src/promise/mod.rs | sed -n '1,260p'
printf '%s\n' ''
printf '%s\n' '--- direct_message imports and start call paths ---'
sed -n '1,110p' crates/perry-stdlib/src/worker_threads/direct_message.rs
rg -n -C 6 'js_worker_threads_post_message_to_thread|wait_for_direct_message_ack|DirectMessaging|worker_messaging_error_value|queue_worker_messaging_rejection' crates/perry-stdlib/src/worker_threadsRepository: PerryTS/perry
Length of output: 50371
Do not run deferred promise converters on background threads.
js_worker_threads_post_message_to_thread() spawns wait_for_direct_message_ack(), and rejection converts worker_messaging_error_value(...) through queue_deferred_resolution(...). In the fallback shim, that converted js_error_new() runs on the spawned thread; in the async bridge, deferred converters are queued because perry-runtime uses thread-local arenas and the async bridge converters create JS values. Keep main-thread conversion in both shims/bridges so rejected promises do not allocate through the background thread’s runtime context.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@crates/perry-stdlib/src/worker_threads/async_shim.rs` around lines 52 - 73,
Update queue_deferred_resolution and the corresponding async bridge so converter
is never executed on the spawned wait_for_direct_message_ack thread. Dispatch
the conversion to the main-thread/runtime context, then call
queue_promise_resolution with the converted value there; preserve the existing
success/rejection behavior and call-site compatibility.
Source: MCP tools
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.
Fixes #7764.
What was broken
cargo build -p perry-stdlib --no-default-featuresproduced 12 errors. That is the configuration the auto-optimize relink builds with, so everyperrycompile that triggered auto-optimize fell back to the prebuilt archives with a warning, and any ad-hoc build neededPERRY_NO_AUTO_OPTIMIZE=1.Both causes violate the contract
common/mod.rsalready states in prose:Cause 1 — one line, exactly as diagnosed (#7745)
js_set_native_events_dispatchreferencedcrate::eventswithout the#[cfg(feature = "bundled-events")]that gates the module. The neighbouring registrations in the same function are gated —database-sqliteon the very next line — which is what makes this an omission rather than a decision. One#[cfg].Cause 2 — eleven sites the issue listed as "the same shape"
They are the same shape but not the same fix.
worker_threadsis always-on and reachescommon::async_bridgein eleven places across five files, and neither obvious repair works:#[cfg]per call site — two of them are value-producing (js_promise_new_for_native_resolution), so gating the statement leaves nothing to return.worker_threadsmodule — worse. It has no feature of its own, so its FFI symbols would vanish from the stripped archive and a program importingnode:worker_threadswould fail to link. That trades a build error for the test_gap_fetch_request_from_node_incoming_message SIGABRTs deterministically on pristine main, and is in no allowlist #7629 family of failure, which is a strictly worse outcome.So
worker_threads/async_shim.rssupplies the four entry points in both configurations — forwarding toasync_bridgewhen it is compiled in, settling inline when it is not.That inline arm is not invented semantics. The queue exists to hand work to the pump; with no pump there is nothing to hand it to, and doing the same work synchronously reaches the same observable end state (the promise settles). The pinning
js_promise_new_for_native_resolutionperforms is likewise a consequence of deferral — it guards the window between creation and the pump's resolution — and an inline settle spans no collection point, so a plainjs_promise_newis its correct counterpart. Each of those arguments is written down at the shim.Validation
cargo build -p perry-stdlib --no-default-features: 0 errors (was 12).cargo build -p perry-stdlib(default features): 0 errors.Both directions deliberately: my first cut of the shim accidentally imported itself, and
--no-default-featurescould not see it because that arm is#[cfg]'d out. Checking only the configuration named in the issue would have shipped it.The actual payoff, checked end to end:
perry compileof a small program now completes the auto-optimize relink with no fallback warning and runs correctly.cargo test -p perry-stdlib --lib: 111 passed, 0 failed.Targeted parity across worker-threads / events / message-port / broadcast / async / promise gap tests: all pass.
cargo fmt --all --check,scripts/check_file_size.sh: clean.One process note: a
test_gap_workerCOMPILE_FAIL appeared once and did not reproduce — it raced a rebuild that was still writing the archives. Re-run twice clean before believing that shape of failure.No version bump.
Summary by CodeRabbit