fix(native): retain processor, hidden-slice and property type facts

Separate AMDGPU processors from feature-qualified note identities. Read Apple hidden ARM64 table rows once and preserve visibility through artifact, dylib and fuse consumers. Keep Objective-C++ template commas within type attributes and incomplete metadata unknown.

Validated provider/parser and production property regressions, Electron CLI/MCP consumers, and native Apple lipo -hideARM64 output through the production slice reader and public artifact inspection. LLVM execution remains unverified.

Fixes #6806

Fixes #6799

Fixes #6801
This commit is contained in:
morluto
2026-10-24 01:00:14 +08:00
parent 6fdcac6cad
commit d4415444c4
24 changed files with 670 additions and 58 deletions
+6 -1
View File
@@ -116,7 +116,12 @@ Kernel representations have different meanings:
including V2 kernel-parameter sizes, and native section declarations. Global
register declarations also cover compiler
versions that leave the older text-section register field empty.
- AMDGPU reads v3+ MessagePack notes. Argument counts include hidden ABI arguments
- AMDGPU reads v3+ MessagePack notes. V3 processor identity comes from the
`EF_AMDGPU_MACH` field in ELF `e_flags`; known processor values are selectable
even when the note omits `amdhsa.target`. This inventory reports the processor
only; it does not interpret version-specific feature flags. Unknown processor values remain unknown. A known
processor conflict between ELF flags and note metadata leaves the processor
unresolved. Argument counts include hidden ABI arguments
when present; SGPR/VGPR counts and fixed group/private segment sizes retain
their declared meanings. Legacy YAML metadata remains unsupported.