Repository navigation
Use RealizedTy in runtime - #3998
Merged
Merged
Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
⏭️ Performance benchmarks were skippedPerf benchmarks (CodSpeed) are opt-in on pull requests — they no longer run on every push. They always run automatically after merge to To run them on this PR, do any of the following, then push a commit (or re-run CI):
|
Contributor
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: Repository UI Review profile: CHILL Plan: Pro Run ID: ⛔ Files ignored due to path filters (1)
📒 Files selected for processing (53)
🚧 Files skipped from review as they are similar to previous changes (49)
📝 WalkthroughWalkthroughThis PR makes associated-type projection interfaces mandatory and migrates compiler, VM, heap, serialization, and host-boundary type metadata from ChangesType-system projection invariants
Realized type runtime migration
Estimated code review effort: 5 (Critical) | ~120 minutes Possibly related PRs
Suggested reviewers: Poem
🚥 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 |
Binary size checks passed✅ 7 passed
Generated by |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
Actionable comments posted: 4
🧹 Nitpick comments (3)
baml_language/crates/bex_vm/src/package_baml/resolve.rs (1)
731-738: 📐 Maintainability & Code Quality | 🔵 Trivial | 💤 Low valueRedundant double
try_fromconversion.
<&RealizedTy>::try_from(other)is evaluated once in the match guard and again inside the arm body. Consider binding it once.♻️ Suggested simplification
- other if <&RealizedTy>::try_from(other).is_ok() => { - let realized = <&RealizedTy>::try_from(other) - .unwrap_or_else(|_| unreachable!("guarded by the `is_ok` above")); - let (args, assoc) = pinned_interface_instantiation(vm, rule); - push_unique(&mut out, (realized.clone(), args, assoc)); - } + other if let Ok(realized) = <&RealizedTy>::try_from(other) => { + let (args, assoc) = pinned_interface_instantiation(vm, rule); + push_unique(&mut out, (realized.clone(), args, assoc)); + }Note: match-guard
letbindings require a sufficiently recent Rust edition/feature; verify compatibility with the project's toolchain before applying, or use anif let/matchrestructure instead.🤖 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 `@baml_language/crates/bex_vm/src/package_baml/resolve.rs` around lines 731 - 738, Update the concrete-type branch in the surrounding resolver logic to perform the RealizedTy conversion only once, binding and reusing the successful result instead of calling try_from in both the guard and arm body. If match-guard bindings are unsupported by the project toolchain, restructure this branch with if let or match while preserving the existing push_unique behavior.baml_language/crates/bex_engine/src/conversion.rs (1)
189-223: 🚀 Performance & Scalability | 🔵 Trivial | ⚡ Quick winHoist the repeated
class_type_args→Vec<RuntimeTy>conversion.The same widening of
instance.class_type_argsis recomputed inside the per-field closure (forfield_type) and again for the outerInstance.type_args, redoing the conversion once per field instead of once per instance.♻️ Suggested refactor
+ let runtime_type_args: Vec<baml_type::RuntimeTy> = instance + .class_type_args + .iter() + .map(baml_type::RuntimeTy::from) + .collect(); let fields: Result<indexmap::IndexMap<String, BexExternalValue>, EngineError> = class .fields .iter() .zip(instance.fields.iter()) .map(|(class_field, slot)| { let value = slot.load(); - let field_type = class_field.field_template.substitute_symbolic( - &instance - .class_type_args - .iter() - .map(baml_type::RuntimeTy::from) - .collect::<Vec<_>>(), - ); + let field_type = class_field + .field_template + .substitute_symbolic(&runtime_type_args); Ok(( class_field.name.clone(), self.convert_vm_value_to_external_with_type( value, &field_type, permit, )?, )) }) .collect(); Ok(BexExternalValue::Instance { class_name: class.name.to_string(), - type_args: instance - .class_type_args - .iter() - .map(baml_type::RuntimeTy::from) - .collect(), + type_args: runtime_type_args, fields: fields?, })🤖 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 `@baml_language/crates/bex_engine/src/conversion.rs` around lines 189 - 223, In the instance conversion branch, hoist the conversion of instance.class_type_args to Vec<RuntimeTy> into a local value before the fields iterator. Reuse that value for substitute_symbolic in the per-field closure and for the Instance type_args field, removing both repeated conversions while preserving the existing field and instance output.baml_language/crates/bex_vm_types/src/types/function.rs (1)
337-360: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick winStale doc comment:
BoundMethod::type_argscomment still says it staysRuntimeTy.The comment claims "
RuntimeTy(notRealizedTy) ... so all three stayRuntimeTyand narrow toRealizedTytogether once it lands" — but this very diff narrowsClosure::captured_type_args,BoundMethod::type_args, andGenericFunction::type_argstoRealizedTyalready. The comment now directly contradicts the field's declared type just below it.📝 Proposed doc update
- /// `RuntimeTy` (not `RealizedTy`) mirrors [`Closure::captured_type_args`] - /// and [`GenericFunction::type_args`]: these positions should never carry a - /// type variable, but the upstream fix that stops typevars leaking into - /// value positions is still in flight, so all three stay `RuntimeTy` and - /// narrow to `RealizedTy` together once it lands. + /// `RealizedTy` (like [`Closure::captured_type_args`] and + /// [`GenericFunction::type_args`]): these positions never carry a type + /// variable, so they are fully realized by construction. pub type_args: Box<[baml_type::RealizedTy]>,🤖 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 `@baml_language/crates/bex_vm_types/src/types/function.rs` around lines 337 - 360, Update the BoundMethod::type_args documentation to reflect that the field and corresponding Closure::captured_type_args and GenericFunction::type_args now use RealizedTy. Remove the stale RuntimeTy explanation and any claim that a future narrowing is still pending.
🤖 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 `@baml_language/crates/bex_vm_types/src/types/class.rs`:
- Around line 80-83: The doc comment above Instance::class_type_args references
the obsolete RuntimeTy type; update it to refer to the current
baml_type::RealizedTy element type while preserving the explanation about the
boxed slice and existing convention.
In `@baml_language/crates/bex_vm_types/src/types/object.rs`:
- Around line 124-125: Update the Object::Type payload to preserve the existing
RuntimeTy wire representation instead of baml_type::RealizedTy, matching the
stable serialization used by Program artifacts. If RealizedTy is required
internally, add an explicit versioned conversion or migration path while keeping
legacy bytecode and packed executables deserializable.
In `@baml_language/crates/bex_vm/src/package_baml/ops.rs`:
- Around line 596-604: Update resolve_equals_eq to propagate failures from
resolve::realize_frame instead of calling unwrap_or_else with unreachable!. Use
the function’s existing Option return path so realization errors produce None,
while preserving the successful Some((callee, type_args)) behavior.
In `@baml_language/crates/bex_vm/src/type_context.rs`:
- Around line 116-158: The runtime associated-type reduction path can recurse
indefinitely through RuntimeTypeContext::project and TyTemplate::substitute. Add
and propagate a projection depth or fuel limit across both
baml_language/crates/bex_vm/src/type_context.rs:116-158 and
baml_language/crates/baml_type/src/template.rs:237-258, terminating with an
opaque or otherwise safe result when exhausted, consistent with
PROJECTION_REDUCTION_FUEL.
---
Nitpick comments:
In `@baml_language/crates/bex_engine/src/conversion.rs`:
- Around line 189-223: In the instance conversion branch, hoist the conversion
of instance.class_type_args to Vec<RuntimeTy> into a local value before the
fields iterator. Reuse that value for substitute_symbolic in the per-field
closure and for the Instance type_args field, removing both repeated conversions
while preserving the existing field and instance output.
In `@baml_language/crates/bex_vm_types/src/types/function.rs`:
- Around line 337-360: Update the BoundMethod::type_args documentation to
reflect that the field and corresponding Closure::captured_type_args and
GenericFunction::type_args now use RealizedTy. Remove the stale RuntimeTy
explanation and any claim that a future narrowing is still pending.
In `@baml_language/crates/bex_vm/src/package_baml/resolve.rs`:
- Around line 731-738: Update the concrete-type branch in the surrounding
resolver logic to perform the RealizedTy conversion only once, binding and
reusing the successful result instead of calling try_from in both the guard and
arm body. If match-guard bindings are unsupported by the project toolchain,
restructure this branch with if let or match while preserving the existing
push_unique behavior.
🪄 Autofix (Beta)
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: Repository UI
Review profile: CHILL
Plan: Pro
Run ID: 3a73d925-7cf2-4d85-8adf-ef303b839053
📒 Files selected for processing (52)
baml_language/crates/baml_builtins2_codegen/src/codegen.rsbaml_language/crates/baml_compiler2_emit/src/lib.rsbaml_language/crates/baml_compiler2_mir/src/ir.rsbaml_language/crates/baml_compiler2_mir/src/lower.rsbaml_language/crates/baml_compiler2_tir/src/builder.rsbaml_language/crates/baml_compiler2_tir/src/builder/associated_projection.rsbaml_language/crates/baml_compiler2_tir/src/exhaustiveness.rsbaml_language/crates/baml_compiler2_tir/src/generics.rsbaml_language/crates/baml_compiler2_tir/src/interfaces.rsbaml_language/crates/baml_compiler2_tir/src/interfaces/coherence.rsbaml_language/crates/baml_compiler2_tir/src/interfaces/impl_rules.rsbaml_language/crates/baml_compiler2_tir/src/lower_type_expr.rsbaml_language/crates/baml_compiler2_tir/src/normalize.rsbaml_language/crates/baml_type/src/family.rsbaml_language/crates/baml_type/src/lib.rsbaml_language/crates/baml_type/src/normalize.rsbaml_language/crates/baml_type/src/normalize/tests.rsbaml_language/crates/baml_type/src/realized_ty.rsbaml_language/crates/baml_type/src/runtime_ty.rsbaml_language/crates/baml_type/src/template.rsbaml_language/crates/bex_engine/src/conversion.rsbaml_language/crates/bex_engine/src/lib.rsbaml_language/crates/bex_engine/src/trace_heap.rsbaml_language/crates/bex_engine/src/trace_value_encode.rsbaml_language/crates/bex_heap/src/accessor.rsbaml_language/crates/bex_heap/src/gc.rsbaml_language/crates/bex_heap/src/tlab.rsbaml_language/crates/bex_heap/tests/generational.rsbaml_language/crates/bex_vm/src/package_baml/array.rsbaml_language/crates/bex_vm/src/package_baml/csv.rsbaml_language/crates/bex_vm/src/package_baml/json.rsbaml_language/crates/bex_vm/src/package_baml/mod.rsbaml_language/crates/bex_vm/src/package_baml/ops.rsbaml_language/crates/bex_vm/src/package_baml/resolve.rsbaml_language/crates/bex_vm/src/package_baml/root.rsbaml_language/crates/bex_vm/src/package_baml/toml.rsbaml_language/crates/bex_vm/src/package_baml/type_class.rsbaml_language/crates/bex_vm/src/package_baml/yaml.rsbaml_language/crates/bex_vm/src/type_context.rsbaml_language/crates/bex_vm/src/type_match.rsbaml_language/crates/bex_vm/src/vm.rsbaml_language/crates/bex_vm/src/watch.rsbaml_language/crates/bex_vm/tests/load_type.rsbaml_language/crates/bex_vm/tests/method_class_type_args.rsbaml_language/crates/bex_vm_types/src/errors.rsbaml_language/crates/bex_vm_types/src/types.rsbaml_language/crates/bex_vm_types/src/types/class.rsbaml_language/crates/bex_vm_types/src/types/containers.rsbaml_language/crates/bex_vm_types/src/types/function.rsbaml_language/crates/bex_vm_types/src/types/object.rsbaml_language/crates/bridge_ctypes/src/ty_decode.rsbaml_language/crates/bridge_ctypes/src/ty_encode.rs
2kai2kai2
force-pushed
the
kai/realized-ty
branch
from
July 14, 2026 01:15
3d3abec to
d6044fa
Compare
2kai2kai2
enabled auto-merge
July 14, 2026 01:15
Contributor
There was a problem hiding this comment.
🧹 Nitpick comments (1)
baml_language/crates/baml_builtins2_codegen/src/codegen.rs (1)
1827-1830: 📐 Maintainability & Code Quality | 🔵 Trivial | 💤 Low valueStale doc: the "absent from
generics" case no longer yieldsunknown. The header still saysunknownappears "when a generic name is absent fromgenerics", but theGenericarm now emitscompile_error!for that case (Line 1868); only the "type argument not supplied at the call" case falls back tounknown. Update the sentence to match.🤖 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 `@baml_language/crates/baml_builtins2_codegen/src/codegen.rs` around lines 1827 - 1830, Update the documentation comment above the relevant generic type-resolution logic to remove the claim that an absent generic name yields unknown. State that unknown is used only when a generic type argument was not supplied at the call, while the absent-from-generics case emits compile_error! as implemented in the Generic arm.
🤖 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.
Nitpick comments:
In `@baml_language/crates/baml_builtins2_codegen/src/codegen.rs`:
- Around line 1827-1830: Update the documentation comment above the relevant
generic type-resolution logic to remove the claim that an absent generic name
yields unknown. State that unknown is used only when a generic type argument was
not supplied at the call, while the absent-from-generics case emits
compile_error! as implemented in the Generic arm.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Repository UI
Review profile: CHILL
Plan: Pro
Run ID: 3f92ccff-7d36-4952-87ef-afcab77ae924
⛔ Files ignored due to path filters (1)
baml_language/crates/baml_tests/snapshots/baml_src/inferred_generic_type_args.snapis excluded by!**/*.snap
📒 Files selected for processing (54)
baml_language/crates/baml_builtins2_codegen/src/codegen.rsbaml_language/crates/baml_compiler2_emit/src/lib.rsbaml_language/crates/baml_compiler2_mir/src/ir.rsbaml_language/crates/baml_compiler2_mir/src/lower.rsbaml_language/crates/baml_compiler2_tir/src/builder.rsbaml_language/crates/baml_compiler2_tir/src/builder/associated_projection.rsbaml_language/crates/baml_compiler2_tir/src/exhaustiveness.rsbaml_language/crates/baml_compiler2_tir/src/generics.rsbaml_language/crates/baml_compiler2_tir/src/interfaces.rsbaml_language/crates/baml_compiler2_tir/src/interfaces/coherence.rsbaml_language/crates/baml_compiler2_tir/src/interfaces/impl_rules.rsbaml_language/crates/baml_compiler2_tir/src/lower_type_expr.rsbaml_language/crates/baml_compiler2_tir/src/normalize.rsbaml_language/crates/baml_compiler2_tir/src/type_context.rsbaml_language/crates/baml_type/src/family.rsbaml_language/crates/baml_type/src/lib.rsbaml_language/crates/baml_type/src/normalize.rsbaml_language/crates/baml_type/src/normalize/tests.rsbaml_language/crates/baml_type/src/realized_ty.rsbaml_language/crates/baml_type/src/runtime_ty.rsbaml_language/crates/baml_type/src/template.rsbaml_language/crates/bex_engine/src/conversion.rsbaml_language/crates/bex_engine/src/future.rsbaml_language/crates/bex_engine/src/lib.rsbaml_language/crates/bex_engine/src/trace_heap.rsbaml_language/crates/bex_engine/src/trace_value_encode.rsbaml_language/crates/bex_heap/src/accessor.rsbaml_language/crates/bex_heap/src/gc.rsbaml_language/crates/bex_heap/src/tlab.rsbaml_language/crates/bex_heap/tests/generational.rsbaml_language/crates/bex_vm/src/package_baml/array.rsbaml_language/crates/bex_vm/src/package_baml/csv.rsbaml_language/crates/bex_vm/src/package_baml/json.rsbaml_language/crates/bex_vm/src/package_baml/mod.rsbaml_language/crates/bex_vm/src/package_baml/ops.rsbaml_language/crates/bex_vm/src/package_baml/resolve.rsbaml_language/crates/bex_vm/src/package_baml/root.rsbaml_language/crates/bex_vm/src/package_baml/toml.rsbaml_language/crates/bex_vm/src/package_baml/type_class.rsbaml_language/crates/bex_vm/src/package_baml/yaml.rsbaml_language/crates/bex_vm/src/type_context.rsbaml_language/crates/bex_vm/src/type_match.rsbaml_language/crates/bex_vm/src/vm.rsbaml_language/crates/bex_vm/src/watch.rsbaml_language/crates/bex_vm/tests/load_type.rsbaml_language/crates/bex_vm/tests/method_class_type_args.rsbaml_language/crates/bex_vm_types/src/errors.rsbaml_language/crates/bex_vm_types/src/types.rsbaml_language/crates/bex_vm_types/src/types/class.rsbaml_language/crates/bex_vm_types/src/types/containers.rsbaml_language/crates/bex_vm_types/src/types/function.rsbaml_language/crates/bex_vm_types/src/types/object.rsbaml_language/crates/bridge_ctypes/src/ty_decode.rsbaml_language/crates/bridge_ctypes/src/ty_encode.rs
🚧 Files skipped from review as they are similar to previous changes (48)
- baml_language/crates/bex_vm/src/package_baml/yaml.rs
- baml_language/crates/baml_compiler2_tir/src/builder.rs
- baml_language/crates/baml_compiler2_tir/src/exhaustiveness.rs
- baml_language/crates/bex_vm/src/package_baml/toml.rs
- baml_language/crates/baml_compiler2_mir/src/ir.rs
- baml_language/crates/baml_type/src/family.rs
- baml_language/crates/bex_vm/src/package_baml/root.rs
- baml_language/crates/bex_vm/src/watch.rs
- baml_language/crates/bex_engine/src/trace_value_encode.rs
- baml_language/crates/bex_vm_types/src/types/containers.rs
- baml_language/crates/bex_vm_types/src/types.rs
- baml_language/crates/baml_compiler2_tir/src/interfaces.rs
- baml_language/crates/bridge_ctypes/src/ty_encode.rs
- baml_language/crates/baml_compiler2_emit/src/lib.rs
- baml_language/crates/bex_engine/src/trace_heap.rs
- baml_language/crates/bex_vm_types/src/types/class.rs
- baml_language/crates/bex_vm/src/package_baml/type_class.rs
- baml_language/crates/bex_vm/src/type_context.rs
- baml_language/crates/bridge_ctypes/src/ty_decode.rs
- baml_language/crates/bex_vm/src/type_match.rs
- baml_language/crates/baml_compiler2_tir/src/interfaces/impl_rules.rs
- baml_language/crates/bex_vm_types/src/errors.rs
- baml_language/crates/baml_compiler2_tir/src/builder/associated_projection.rs
- baml_language/crates/bex_vm_types/src/types/object.rs
- baml_language/crates/baml_compiler2_tir/src/generics.rs
- baml_language/crates/baml_type/src/lib.rs
- baml_language/crates/baml_type/src/realized_ty.rs
- baml_language/crates/baml_compiler2_tir/src/interfaces/coherence.rs
- baml_language/crates/baml_compiler2_tir/src/type_context.rs
- baml_language/crates/baml_type/src/runtime_ty.rs
- baml_language/crates/bex_vm/tests/method_class_type_args.rs
- baml_language/crates/bex_heap/src/accessor.rs
- baml_language/crates/bex_vm/src/package_baml/mod.rs
- baml_language/crates/bex_vm_types/src/types/function.rs
- baml_language/crates/baml_compiler2_tir/src/normalize.rs
- baml_language/crates/bex_heap/src/gc.rs
- baml_language/crates/bex_heap/tests/generational.rs
- baml_language/crates/bex_heap/src/tlab.rs
- baml_language/crates/bex_vm/src/package_baml/csv.rs
- baml_language/crates/baml_compiler2_mir/src/lower.rs
- baml_language/crates/bex_vm/tests/load_type.rs
- baml_language/crates/baml_type/src/normalize/tests.rs
- baml_language/crates/baml_type/src/normalize.rs
- baml_language/crates/baml_compiler2_tir/src/lower_type_expr.rs
- baml_language/crates/bex_engine/src/conversion.rs
- baml_language/crates/baml_type/src/template.rs
- baml_language/crates/bex_vm/src/package_baml/ops.rs
- baml_language/crates/bex_vm/src/package_baml/resolve.rs
github-merge-queue
Bot
removed this pull request from the merge queue due to failed status checks
Jul 14, 2026
A thread that terminated with an internal error is now killed instead of hanging forever. Also placed a bandaid on the case that fixed it (broader fix planned for proper dynamic `Self` resolution in default interface method bodies)
2kai2kai2
force-pushed
the
kai/realized-ty
branch
from
July 14, 2026 21:37
d6044fa to
d3c41ae
Compare
2kai2kai2
enabled auto-merge
July 14, 2026 21:45
antoniosarosi
added a commit
that referenced
this pull request
Jul 15, 2026
Semantic reconcile with canary's #3998 ("Use RealizedTy in runtime"). Canary's change: the runtime narrowed every type-arg storage slot from RuntimeTy to RealizedTy (the typevar-free deep subset of the ty_family) — Instance::class_type_args, Array::element_ty, Map::key_ty/value_ty, Closure::captured_type_args, BoundMethod/GenericFunction::type_args, HostClosure::ret_ty/throws_ty/params, Value::{Type,Array,Map}, plus the VM's BytecodeFrame::type_args / pending_call_type_args and accessors. (It also made AssociatedTypeProjection::interface non-optional, but that surface isn't touched by this branch.) The merge was textually clean; the break was purely a type mismatch at our call sites (Vec<RuntimeTy> vs the new &[RealizedTy]). Our adaptation — mirror the same narrowing at the bytecode-cache/linker sites that shadow GenericFunction::type_args: - GenericFnKey::type_args: Vec<RuntimeTy> -> Vec<RealizedTy> (unit.rs) - linker generic-fn dedup map: HashMap<String, HashMap<Vec<RuntimeTy>, usize>> -> Vec<RealizedTy> (link.rs, both the resolver and the interning pass) - emit's tail_generic_dupes_clean clean_keys: Vec<(String, Vec<RuntimeTy>)> -> Vec<RealizedTy> (lib.rs) This is a pure rename/narrow: the generic-value interning model is unchanged (intern key is still (base_fn, type_args)); only the element type narrows. Borsh wire shape: GenericFnKey::type_args serializes a RealizedTy instead of a RuntimeTy, but ty_family assigns each variant its explicit stable master discriminant (RealizedTy just omits the typevar/projection variants, leaving gaps), so any realized value's on-wire bytes are identical. Even setting that aside, FORMAT_VERSION is 1 and unreleased, so a shape change would be fine. Gate: cargo check --workspace clean (0 warnings); fmt/clippy clean; baml_cli --lib (358), bex_vm_types+bex_cache+bex_project (106), relink_oracle (1), link_units_oracle (5), bytecode_format (7, no snapshot regen needed) all green. BAML_CACHE_VERIFY=1 corpus spot-check (cold + warm + body-edit reuse/relink, "wrote 1, reused 76") zero bails.
antoniosarosi
added a commit
that referenced
this pull request
Jul 15, 2026
… Windows size ceiling Two CI-only failures on non-Linux platforms. 1. diagnostics_cache_oracle: the `unique_root()` helper anchored the project root under `std::env::temp_dir()` verbatim (unlike the already-fixed `bytecode_cache.rs` copy). `ProjectDatabase` canonicalizes the root and source paths only when they exist on disk: db1 is built before the cache dir exists (root canonicalize is a no-op fallback), then `store_with_manifest` materializes the root, so db2's `set_project_root` canonicalizes it — while the in-memory `.baml` files never exist to canonicalize. On macOS the root then resolves `/var` -> `/private/var`, on Windows it gains the `\\?\` prefix; the source paths keep the unresolved form, `strip_prefix` fails, every rel_path comes out absolute, and `plan_reuse` collapses to `None` — so `plan.dirty_files` is empty. `served == honest` still holds (a `None` plan is a full fresh check), but the `dirty.contains(...)` assertions fail (oracle_layout_reorder, oracle_mixed_file_field_reorder, oracle_throws_edit, verify_diagnostics_bails_on_a_stale_cache). `/tmp` on Linux has no symlink so it was silently fine. Canonicalize the temp base up front, matching 4f6df7b. Reproduced on Linux by routing TMPDIR through a symlink (same 4 failures); fixed under the symlink and in a normal run. 2. Size gate: the Windows baml-cli absolute ceiling was miscalibrated at baseline x1.0199 (19_098_870 / 18_726_400) — tighter than the 3% max_delta_pct guard and every other platform's x1.03 ceiling. Post-baseline canary merges (#3998 RealizedTy runtime, #4031 array-rest patterns; no Cargo.lock change) grew the Windows binary +408.6 KB (+2.18%, within the 3% guard) to 19_134_976, tripping only the stale ceiling. Normalize it to baseline x1.03 = 19_288_192, matching the 68f72e2 precedent.
This was referenced Jul 15, 2026
This was referenced Jul 27, 2026
This was referenced Aug 6, 2026
This was referenced Aug 24, 2026
This branch was previously deployed
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
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.
Representation tightening to prevent invariant violations:
Iin(C as I).Assoc) is no longer optional: it must either be written explicitly or inferred by the TIR.RealizedTyanywhere that typevars are invalid, including function call frames and instances' generic params.Summary by CodeRabbit
Bug Fixes
Reliability