fix(types): skip ForwardRef resolution for unannotated attribute probes

[benedikt-bartscher]

[greptile-apps]

Greptile Summary

This PR avoids resolving unrelated forward references during unannotated attribute probes and consolidates event callback type-hint resolution.

• Adds annotation-name probing across a class MRO.
• Gates full type-hint resolution on whether the requested attribute is annotated.
• Adds regression coverage for unresolved and Python 3.14 lazy annotations.
• Extracts the event callback type-hint fallback into a shared helper.

Confidence Score: 4/5

The PR is not yet safe to merge because the annotation-name cache can reject attributes added after its first snapshot.

The cached MRO annotation-name set is never refreshed, so a later class annotation remains invisible to the new gate and ObjectVar attribute access raises VarAttributeError.

Files Needing Attention: packages/reflex-base/src/reflex_base/utils/compat.py and packages/reflex-base/src/reflex_base/utils/types.py

Important Files Changed

Filename	Overview
packages/reflex-base/src/reflex_base/utils/compat.py	Adds cached MRO annotation-name discovery, but the cache remains stale after class annotation mutation.
packages/reflex-base/src/reflex_base/utils/types.py	Uses annotation-name discovery to avoid unnecessary ForwardRef resolution, while relying on the potentially stale cached result.
packages/reflex-base/src/reflex_base/event/init.py	Extracts the existing NameError fallback for callback type-hint resolution into a shared helper.
tests/units/test_attribute_access_type.py	Covers unannotated probes, unresolved annotations, and Python 3.14 lazy annotations, but does not cover annotation mutation after an initial probe.

_{Reviews (2): Last reviewed commit: "fix: actually test it for python3.14" | Re-trigger Greptile}

[codspeed-hq]

Merging this PR will not alter performance

✅ 27 untouched benchmarks
⏩ 8 skipped benchmarks¹

---

_{Comparing benedikt-bartscher:annotation-probe-guard (a6b988d) with main (d86f167)}

Footnotes

1. 8 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[cubic-dev-ai]

All reported issues were addressed across 5 files

<sub>Reply with feedback, questions, or to request a fix.

Re-trigger cubic</sub>