Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
32 changes: 31 additions & 1 deletion .github/workflows/cache-warm.yml
Original file line number Diff line number Diff line change
Expand Up @@ -41,7 +41,10 @@ jobs:
# Best-effort cache population — must never block or fail anything.
continue-on-error: true
# Comfortably above the cold first run; later runs are far shorter.
timeout-minutes: 120
# 150: the release-compiler + gap-suite variant warm below adds up to
# ~60 min on a cold variant set (it is the same bill the smoke job was
# paying per-PR; here it is paid once per merge, off the critical path).
timeout-minutes: 150
env:
RUSTC_WRAPPER: sccache
# Mirror test.yml's sccache config EXACTLY (local disk cache, NOT the GHA
Expand Down Expand Up @@ -97,6 +100,33 @@ jobs:
find target/debug/deps -maxdepth 1 -type f -perm -111 ! -name '*.so' -delete || true
done

# Warm the RELEASE compiler + the auto-optimize archive variants for the
# conformance-smoke job. Diagnosis (2026-07-04, cancelled smoke job
# 85123031828): smoke restored this cache with a FULL match yet still
# timed out at 60 min — the debug units above share nothing with its
# release build (~10 min), and every gap test whose feature combo
# (`auto_optimized_cache_key`: regex/temporal/url/… × stdlib set) had no
# `target/perry-auto-<hash>` archive triggered an in-job cargo build of
# the runtime — orphan `cargo`/`rustc` processes were still compiling at
# the timeout. Running the gap suite ONCE here builds every variant the
# suite needs; they land in `target/perry-auto-*`, which rust-cache
# saves under the shared key. Variant keys embed CARGO_PKG_VERSION, and
# code-only PRs carry main's version, so PR smokes restore exact
# matches until the next merge bumps the version — at which point THIS
# workflow re-warms on that very push. `|| true`: this is a cache warm,
# not a gate — the smoke job is where gap failures are reported.
- name: Warm release compiler + gap-suite auto-opt variants
run: |
echo "::group::cargo build --release -p perry"
cargo build --release -p perry
echo "::endgroup::"
echo "::group::run_gap_tests (variant warm)"
./scripts/run_gap_tests.sh || true
echo "::endgroup::"
echo "perry-auto variants warmed:"
du -sh target/perry-auto-* 2>/dev/null || echo "(none)"
df -h /home/runner 2>/dev/null | tail -1 || true

- name: sccache stats
if: always()
run: sccache --show-stats || true
6 changes: 6 additions & 0 deletions CHANGELOG.md
Original file line number Diff line number Diff line change
@@ -1,3 +1,9 @@
## v0.5.1233 — CI: cache-warm primes the release compiler + gap-suite auto-opt variants

The conformance-smoke job restored the shared rust-cache with a full match and still hit its 60-minute timeout: the warmed DEBUG units share nothing with its RELEASE build, and every gap test whose auto-optimize feature combination had no `target/perry-auto-<hash>` archive kicked off an in-job cargo build of the runtime (orphan cargo/rustc still compiling at cancellation). cache-warm now builds the release compiler and runs the gap suite once per main merge so every variant lands under the shared key; code-only PRs carry main's `CARGO_PKG_VERSION`, so their smoke jobs restore exact variant matches until the next merge re-warms (job timeout 120 → 150 for the cold first run).

Also brings the version stream back in sync with two code-only fleet merges that landed without metadata: #5953 (#5898 — mutating array methods throw on frozen/non-writable length) and #5954 (#5909 — `JSON.stringify` of a toJSON-returned cycle throws instead of SIGSEGV).

## v0.5.1232 — anchor inliner + deforest max-LocalId scans on the hardened id_scan

The inliner (`inline/analysis.rs`) and deforestation (`deforest/walk.rs`) each carried their own ad-hoc max-LocalId scan to mint fresh locals; both were shallower than `generator::compute_max_local_id` (the scan hardened for #5143 to cover class field/static/computed-member/extends closures). A fresh local minted from an under-counting scan can collide with a LocalId owned by a closure the scan missed — the same defect class as #5143's FuncId collision. Both passes now seed from `crate::generator::compute_max_local_id(module)` and only extend it with their local maxima, so every future descent fix to the shared scan protects all three passes at once.
Expand Down
2 changes: 1 addition & 1 deletion CLAUDE.md
Original file line number Diff line number Diff line change
Expand Up @@ -8,7 +8,7 @@ This file provides guidance to Claude Code (claude.ai/code) when working with co

Perry is a native TypeScript compiler written in Rust that compiles TypeScript source code directly to native executables. It uses SWC for TypeScript parsing and LLVM for code generation.

**Current Version:** 0.5.1232
**Current Version:** 0.5.1233


## TypeScript Parity Status
Expand Down
Loading
Loading