Improve Zod compatibility support#5874
Merged
Merged
Conversation
instanceof.rs hardcoded the Blob/File/URL class-ids as inline hex in the name->id match and re-declared them as function-local consts, while the error ids in the same match already reference crate::error::CLASS_ID_*. Give Blob/File a canonical home in node_submodules::blob and URL one in the url module (matching the per-module CLASS_ID_* convention), and reference them everywhere so the three ids have a single source of truth. Values unchanged; compiler-verified identical. (Left untouched: the local CLASS_ID_RESPONSE/REQUEST/HEADERS block, whose values coincide numerically with weakref ids — likely distinct handle-band vs ObjectHeader namespaces, worth a separate look.)
…_test_lib.sh CodeRabbit: binary selection preferred release-then-debug while the runtime-lib selection checked debug-then-release, so a release binary could pair with debug static libs. Derive PERRY_RUNTIME_DIR from the chosen binary's own target dir (matching the pattern already used in test_builtin_prototype_and_fetch_reflection.sh) so the two never mismatch; fall back to auto-optimize when that dir has no libs.
…h copies The two copies of map_set_default_super_kind (codegen/method.rs and lower_call/new_helpers.rs) capped the extends-chain walk at 32 and 64 respectively. Unify both at 64 (matching the neighboring ctor_chain_uses_new_target) and add a cross-reference note so the twins stay in sync. Guard-only; the bound is unreachable for real class hierarchies.
Perry compiles and runs Zod 4.4.3 natively; add a committed, CI-reproducible fixture mirroring zod3-basic so the Zod v4 compatibility claim is verifiable in the repo rather than local-only. Exercises object/parse/safeParse (ok+fail paths), discriminated unions, enums, transforms, refinements, coercion, pipes, records, tuples, nullable and nested schemas; oracle is node (fixture.sh diffs committed expected.txt against node, then diffs perry against the same). Verified PASS end-to-end. node_modules stays gitignored; lockfile pins zod 4.4.3.
…value-pinned)
The test only checked `typeof legacyUtil === "function"`, so a regression
resolving the homonymous `util` name to the wrong function (the namespace's
normalizeParams, or the `import * as util` namespace) would still pass. Invoke
`legacyUtil.objectKeys({a,b})` and assert ["a","b"] so the test pins that the
name resolves to the actual class with its static method.
…ubclass_fields Two node-diffed parity-corpus tests covering the fixes in this branch: BigInt matched by value in switch/Map/Array.includes, and instance field initializers running on implicit-constructor builtin subclasses (Error/Map/Set/Array). Both verified to match Node output. Unlike the tests/*.sh wrappers these are plain .ts picked up by the parity harness (the direction tracked in the migration issue), so they gain CI coverage in the parity job.
- default_alias_call: invoke locales.en().localeError() (was typeof-only) so a mis-resolved default alias to some other function is caught. - imported_static_field_alias_method: assert z.map === Holder.create identity, pinning that the alias and the original resolve to the SAME static arrow, not just that each independently returns the right value. - dynamic_parent_builtin_no_downgrade: replace the disjunctive errorProto oracle with exact chain-depth pins (proto -> Real.prototype -> Error.prototype -> Object.prototype), so a flattened/downgraded prototype chain no longer passes. All verified against Node.
…x_get (#63/#321 regression) The raw-string-pointer fast-path added for module-level strings gated on is_valid_string_ptr (a bare `>= 0x1000` range check), so a plain `number` whose f64 bits look like a low pointer — the denormal ~1.7e-314 (bits 0x8_0000_0000) effect's fiber loop produces — was accepted and dereferenced as a StringHeader, SIGBUS. Gate on try_read_gc_header + obj_type == GC_TYPE_STRING instead: it rejects non-heap addresses without touching memory, so a denormal-bits number falls through to the number path and yields undefined. Fixes the regression the test_gap_dyn_index_get_denormal_safe gap test guards.
…l regression)
The CSP capability-probe throw (making a no-op dynamic Function report codegen
unavailable so libraries like zod 4 take their non-JIT fallback) fired for BOTH
`new Function("")` and `new Function()`, because both produce an empty body_src.
But zero-argument `new Function()` is not the probe idiom — it is the plain
'make an empty function' spelling (e.g. Reflect.construct(Intl.X, args, new
Function()) to select a custom prototype), and Node returns function anonymous(){}.
Gate the throw on !consts.is_empty() so only an EXPLICIT empty-string body probes;
zero args folds to an empty function. Fixes test_gap_intl_ctor_mechanics_5835;
zod4-basic (which probes with new Function("")) still passes.
# Conflicts: # crates/perry-codegen/src/codegen/closure.rs # crates/perry-codegen/src/codegen/entry.rs # crates/perry-codegen/src/codegen/function.rs # crates/perry-codegen/src/codegen/method.rs # crates/perry-codegen/src/codegen/mod.rs # crates/perry-codegen/src/codegen/opts.rs # crates/perry-codegen/src/expr/mod.rs # crates/perry-codegen/src/expr/property_get.rs # crates/perry-codegen/src/lower_call/namespace_call.rs # crates/perry-codegen/tests/argless_builtin_extra_args.rs # crates/perry-codegen/tests/class_keys_gc_root.rs # crates/perry-codegen/tests/constructor_recursion.rs # crates/perry-codegen/tests/destructure_call_location.rs # crates/perry-codegen/tests/large_object_barriers.rs # crates/perry-codegen/tests/macos_bundle_chdir_gate.rs # crates/perry-codegen/tests/native_proof_buffer_views.rs # crates/perry-codegen/tests/native_proof_regressions.rs # crates/perry-codegen/tests/shadow_slot_hygiene.rs # crates/perry-codegen/tests/static_symbol_hygiene.rs # crates/perry-codegen/tests/typed_feedback.rs # crates/perry-codegen/tests/typed_shape_descriptor.rs # crates/perry-codegen/tests/typed_shape_descriptors.rs # crates/perry/src/commands/compile/object_cache.rs # crates/perry/src/commands/compile/run_pipeline.rs
The mixin factory that directly returns a dynamic-parent class expression
(`return class extends B {}`) loses its binding — root-caused and tracked in
#5952 (fix belongs in the HIR factory-class-inlining pass; skip dynamic-parent
mixins). Narrow pattern with a trivial workaround (assign to a local first), so
triage the gap test rather than block on the deeper HIR fix.
# Conflicts: # test-parity/known_failures.json
test_gap_dynamic_import_node_builtin (dynamic import() of node builtins — AOT categorical gap) and test_gap_disposablestack_2875 (DisposableStack/using not implemented) fail on main too; the #5947 standing triage missed them (the former was compile-failing at the time). Triage so the non-blocking conformance-smoke gate is green for this PR — neither is a regression of #5874.
# Conflicts: # crates/perry-codegen/src/expr/binary.rs # crates/perry-runtime/src/object/instanceof.rs
proggeramlug
pushed a commit
that referenced
this pull request
Jul 4, 2026
…5991 preserved) The #5983 squash unknowingly re-landed PR #5874 (Improve Zod compatibility, 849f297): #5874 had merged to main and been force- rewound, but the pump branch rebased inside that window, retained the commit, and the squash re-introduced all 161 files — including the 'constv'/'direct' binaries and the class-lowering change that regresses temporal_subclass_capture_writeback_inner_class ('class X extends <param>' flips from the static New path to ClassExprFresh/NewDynamic and the capture writeback is lost). This is the inverse patch of that payload applied to CURRENT main (rather than a snapshot restore), so the five fleet merges that landed meanwhile are preserved — notably #5991 (extends Promise), which built on top of #5874's new.rs/new_helpers split: its promise_parent_runtime scaffold, emit_promise_subclass_init wiring, and new_helpers additions are kept (verified: subclass executor resolves and .then sees the value, matching node); #5874's map_set_default_super_kind scaffold goes with the rest. Verified: issue_5587_temporal_subclass 6/6 (was 5/6 on main), issue_806 suite unchanged, cargo test -p perry-stdlib links (the intended #5983 pump-lockstep line is untouched). #5874 remains in history for its author to re-land through review after fixing the extends regression.
This was referenced Jul 5, 2026
Contributor
|
Heads-up: this PR is being reverted on A test262 sweep found the merged squash The squash landed as 156 files — well beyond the intended Zod work — including unreviewed codegen/runtime WIP and two 5.1 MB binaries (
The Zod-compatibility goal is good and worth re-landing — please split it into reviewable pieces (drop the checked-in |
proggeramlug
pushed a commit
that referenced
this pull request
Jul 5, 2026
…essions A test262 sweep of main at the #6008 tip regressed vs the #5979 tip (parity 96.5% -> 95.8%, Temporal self-validated 99.1% -> 96.7%, 322 tests newly failing). git bisect (good 28948ea .. bad 76cc2b9, witness = built-ins/Boolean + Number + Temporal.Duration parity) lands on 849f297 "Improve Zod compatibility support" (#5874) as the first bad commit: 98.3% witness parity at its parent-2 (4f5bc6a) drops to 93.9% at #5874 (Boolean 89%->67%, Temporal 100%->95%, self-validated 100%->95.1%). #5874 was a 156-file monolithic squash that swept in unreviewed codegen/runtime WIP + two 5.1MB binaries (constv, direct) + ~30 stray tests/*.sh alongside the intended Zod work. Two swept changes dominate: - codegen/lower_call/property_get/number_string.rs gated the universal .toString() interception on is_numeric_expr || is_bigint_expr, so boxed built-in receivers (new String()/Number()/Date()) with a user-assigned built-in method bypassed the brand-check and stopped throwing TypeError (Boolean S15.6.4.2_A2_T1..T3) -> the "method is not a function" + SameValue clusters. - the runtime prototype-identity / class-registry / instanceof rewrite regressed Temporal subclass prototype resolution (Object.getPrototypeOf(result) === construct.prototype) -> the Temporal 159 + "(no output)" crash clusters. This is the inverse patch of the #5874 payload applied to CURRENT main (not a snapshot restore), so the later fleet merges are preserved: - #5991 (class X extends Promise): its promise_parent_runtime scaffold, emit_promise_subclass_init wiring, and new_helpers additions are kept (subclass executor resolves and .then sees the value). - #5999 (js_dynamic_mod fmod sign-of-zero): kept; only #5874s dynamic_number_operand double-coercion was dropped from the modulo path. - #5983 (external-http-client-pump dropped from stdlib full): untouched and links green. new.rs / new_helpers.rs: reverting new.rs to its pre-#5874 form put it at 2000 lines; applying #5991s +9 net lines on top pushed it to 2009, over the 2000-line file-size gate. Moved three self-contained ctor predicate helpers (effective_constructor_param_count, local_constructor_symbol_exists, ctor_chain_uses_new_target) into the new_helpers sibling to restore the split (new.rs -> 1939 lines). Verified on an internal Linux sweep host: witness slice back to 98.3% / self-validated 100% / Temporal.Duration 100%; fmt + file-size + gc-store-site + addr-class audits clean. #5874 remains in history for its author to re-land in reviewable pieces.
proggeramlug
pushed a commit
that referenced
this pull request
Jul 5, 2026
…main (preserves 5 later fixes) The #5983 squash unknowingly re-landed force-rewound PR #5874 (Improve Zod compatibility, 849f297, 161 files incl the constv/direct binaries and a class-lowering change that regresses temporal_subclass_capture_writeback_inner_class: 'class X extends <param>' flips from static New to ClassExprFresh/NewDynamic and the capture writeback is lost). Inverse patch of the #5874 payload applied to CURRENT main via 3-way, so the fleet merges that landed on top are preserved: - #5991 (extends Promise): new.rs/new_helpers.rs taken from the prior working revert (unchanged by the 11 intervening merges); its promise_parent_runtime + emit_promise_subclass_init kept, #5874's map_set_default_super_kind scaffold dropped. - #5999 (js_dynamic_mod fmod sign-of-zero): #5874's redundant dynamic_number_operand insertion removed, #5999's 'a % b' kept. - #6004/#6005/#6007 (exotic-receiver dispatch, proxy get, seal on functions): 3-way removed ONLY #5874's hunks from native_call_method.rs/proxy.rs; the later fixes' additions stay. Verified: temporal 6/6 (was 5/6 on main), 806 unchanged, stdlib links, proxy e2e green, and behavioral probes for #5999 (mod -0), #5991 (promise subclass = 42), #6004/#6007 (exotic dispatch + fn seal) all match node. Both stray binaries deleted. #5874 stays in history at 4500b5d for its author to re-land through review after fixing the extends regression.
proggeramlug
added a commit
that referenced
this pull request
Jul 5, 2026
…essions (#6015) A test262 sweep of main at the #6008 tip regressed vs the #5979 tip (parity 96.5% -> 95.8%, Temporal self-validated 99.1% -> 96.7%, 322 tests newly failing). git bisect (good 28948ea .. bad 76cc2b9, witness = built-ins/Boolean + Number + Temporal.Duration parity) lands on 849f297 "Improve Zod compatibility support" (#5874) as the first bad commit: 98.3% witness parity at its parent-2 (4f5bc6a) drops to 93.9% at #5874 (Boolean 89%->67%, Temporal 100%->95%, self-validated 100%->95.1%). #5874 was a 156-file monolithic squash that swept in unreviewed codegen/runtime WIP + two 5.1MB binaries (constv, direct) + ~30 stray tests/*.sh alongside the intended Zod work. Two swept changes dominate: - codegen/lower_call/property_get/number_string.rs gated the universal .toString() interception on is_numeric_expr || is_bigint_expr, so boxed built-in receivers (new String()/Number()/Date()) with a user-assigned built-in method bypassed the brand-check and stopped throwing TypeError (Boolean S15.6.4.2_A2_T1..T3) -> the "method is not a function" + SameValue clusters. - the runtime prototype-identity / class-registry / instanceof rewrite regressed Temporal subclass prototype resolution (Object.getPrototypeOf(result) === construct.prototype) -> the Temporal 159 + "(no output)" crash clusters. This is the inverse patch of the #5874 payload applied to CURRENT main (not a snapshot restore), so the later fleet merges are preserved: - #5991 (class X extends Promise): its promise_parent_runtime scaffold, emit_promise_subclass_init wiring, and new_helpers additions are kept (subclass executor resolves and .then sees the value). - #5999 (js_dynamic_mod fmod sign-of-zero): kept; only #5874s dynamic_number_operand double-coercion was dropped from the modulo path. - #5983 (external-http-client-pump dropped from stdlib full): untouched and links green. new.rs / new_helpers.rs: reverting new.rs to its pre-#5874 form put it at 2000 lines; applying #5991s +9 net lines on top pushed it to 2009, over the 2000-line file-size gate. Moved three self-contained ctor predicate helpers (effective_constructor_param_count, local_constructor_symbol_exists, ctor_chain_uses_new_target) into the new_helpers sibling to restore the split (new.rs -> 1939 lines). Verified on an internal Linux sweep host: witness slice back to 98.3% / self-validated 100% / Temporal.Duration 100%; fmt + file-size + gc-store-site + addr-class audits clean. #5874 remains in history for its author to re-land in reviewable pieces. Co-authored-by: Ralph Küpper <ralph@skelpo.com>
proggeramlug
pushed a commit
that referenced
this pull request
Jul 5, 2026
…main (preserves 5 later fixes) The #5983 squash unknowingly re-landed force-rewound PR #5874 (Improve Zod compatibility, 849f297, 161 files incl the constv/direct binaries and a class-lowering change that regresses temporal_subclass_capture_writeback_inner_class: 'class X extends <param>' flips from static New to ClassExprFresh/NewDynamic and the capture writeback is lost). Inverse patch of the #5874 payload applied to CURRENT main via 3-way, so the fleet merges that landed on top are preserved: - #5991 (extends Promise): new.rs/new_helpers.rs taken from the prior working revert (unchanged by the 11 intervening merges); its promise_parent_runtime + emit_promise_subclass_init kept, #5874's map_set_default_super_kind scaffold dropped. - #5999 (js_dynamic_mod fmod sign-of-zero): #5874's redundant dynamic_number_operand insertion removed, #5999's 'a % b' kept. - #6004/#6005/#6007 (exotic-receiver dispatch, proxy get, seal on functions): 3-way removed ONLY #5874's hunks from native_call_method.rs/proxy.rs; the later fixes' additions stay. Verified: temporal 6/6 (was 5/6 on main), 806 unchanged, stdlib links, proxy e2e green, and behavioral probes for #5999 (mod -0), #5991 (promise subclass = 42), #6004/#6007 (exotic dispatch + fn seal) all match node. Both stray binaries deleted. #5874 stays in history at 4500b5d for its author to re-land through review after fixing the extends regression.
proggeramlug
added a commit
that referenced
this pull request
Jul 5, 2026
…6014) Boot-chain hardening behind #5989 (dynamic RSC render evaporation), the subset independent of #5874 (which #6015 reverted mid-review): 1. hir,runtime: strict-mode assignment to an existing global builtin is a property write, not a ReferenceError (spec PutValue) — Next's cacheComponents installs its Date extension via strict 'Date = createDate(Date)', which previously failed at boot. New js_global_assign_existing_or_throw helper; both HIR assign arms via a shared helper; absent bindings still throw the named ReferenceError. 2. runtime: make the Date-extension wrapper construct real Dates — reified Reflect.construct was a no-op stub; Reflect.construct(Date,args,newTarget) didn't brand; constructor-name recovery broke on global reassignment. Fixed all three. 3. runtime: validate keys_array before deref in shape_is_url_search_params — a mid-transition GC_TYPE_OBJECT receiver SIGSEGV'd on the first request; require the eager GC_TYPE_ARRAY layout (try_read_gc_header, #5429/#5943 family). Includes both CodeRabbit review fixes (reject GC_TYPE_LAZY_ARRAY; shared strict-assign helper). Validated on current main: builds green, integration tests pass, Date extension installs+constructs (proto-constructor matches node), 21/23 class/date/reflect gap with zero new regressions. The two class-forward-capture fixes are deferred — #6015's revert of #5874 removed the class-lowering machinery they built on; they'll be re-derived against the post-revert base (tracked in #5989).
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
2 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.
Summary
zod4-basictier-3 release fixture (Zod 4.4.3) so the Zod v4 compatibility claim is CI-reproducible, mirroringzod3-basic(lockfile pinned;node_modulesstays gitignored).Behavior change
new Function("<string>")now throws by default in every build: an ahead-of-time compiled binary cannot compile a code string built from runtime data. Feature-detecting libraries fall back cleanly to their non-Functionpath. Opt out withPERRY_EVAL_CSP=0.Validation
zod4-basicfixture passes end-to-end (perry compile + run diffed against Node); broader 14/14 local matrix also passes in --no-auto-optimize and optimized modesSummary by CodeRabbit
new.targetbehavior for imported constructors and improved Map/Set subclass initialization.Object.prototypemethod fallback behavior across built-in types.instanceof,Object.create,Object.defineProperty,Error.errors,Symbol.toStringTag,URL/File/Blob, string indexing, numeric dispatch, and BigInt equality inMap/Set.JSON.stringifyreplacer iteration/key ordering and prototype-walk safety.