[shims] Move bit masks to SwiftShims and include/swift/ABI. #12306
Conversation
|
@swift-ci please test |
| #define SWIFT_ABI_S390X_SWIFT_SPARE_BITS_MASK 0x0000000000000007ULL | ||
|
|
||
| #endif /* SWIFT_ABI_SYSTEM_H */ | ||
| #endif // __SWIFT_ABI_SYSTEM_H__ |
There was a problem hiding this comment.
Missing newline at end of file
Move bits mask from Metadata.h to SwiftShims's HeapObject.h. This exposes the bit masks to the stdlib, so that the stdlib doesn't have to have its own magic numbers per-platform. This also enhances readability for BridgeObject, whose magic numbers are mostly derived from Swift's ABI.
milseman
force-pushed
the
spooky_bit_masques
branch
from
ee08fc7 to
3d04fb5
Compare
|
@swift-ci please test |
|
Build failed |
|
Build failed |
| /// Corresponding namespaced decls | ||
| #ifdef __cplusplus | ||
| namespace heap_object_abi { | ||
| static const uintptr_t LeastValidPointerValue = |
There was a problem hiding this comment.
Should these be __swift_uintptr_t?
There was a problem hiding this comment.
Actually, what is the point of __swift_uintptr_t? My guess was that __swift_uintptr_t was so that it would be specially imported to the stdlib as UInt rather than UInt<n>. If that's the case, then this isn't visible to the stdlib as it's C++. I'm happy to make the style match, I'm just curious whether there's something more going on here.
edit: formatting and grammar
There was a problem hiding this comment.
The reason is given in SwiftStdint.h. Briefly: it breaks a dependency cycle on Darwin and avoids a tools bug on Linux.
|
@swift-ci please smoke test |
Move bits mask from Metadata.h to SwiftShims's HeapObject.h. This
exposes the bit masks to the stdlib, so that the stdlib doesn't have
to have its own magic numbers per-platform. This also enhances
readability for BridgeObject, whose magic numbers are mostly derived
from Swift's ABI.