closure: make it build on Windows x86_64 #1440
Conversation
|
CC: @parkera |
|
@swift-ci please test |
1 similar comment
|
@swift-ci please test |
|
Seems reasonable. I thought we were actually depending on the swift-corelibs-libdispatch version of this file now, though. |
closure/runtime.c
Outdated
| @@ -142,7 +139,7 @@ GC support stub routines | |||
|
|
|||
|
|
|||
|
|
|||
| static void *_Block_alloc_default(const unsigned long size, const bool initialCountIsOne, const bool isObject) { | |||
| static void *_Block_alloc_default(uintptr_t size, const bool initialCountIsOne, const bool isObject) { | |||
There was a problem hiding this comment.
Shouldn't this still be size_t though? I can see using the ptr types for the CAS.
|
Hmm, actually, yeah, the |
Windows x86_64 is a LLP64 environment. `unsigned long` is not `size_t`. Adjust the parameter to be the correct size for the allocator. This is compatible with the previous implementation on LP64 environments. Prefer to use the `__sync_bool_compare_and_swap` over the MSVC intrinsic which is overly strong. While in the area, remove an unused compare and swap method.
|
Switched to |
|
@compnerd I have been wondering that myself recently as well. We did need this during bringup (dispatch was not yet ready), and I wonder if it is still useful and still working for other platforms that are in that stage. |
|
Like Windows 64 bit, apparently. :) |
|
@swift-ci please test |
1 similar comment
|
@swift-ci please test |
|
That failure is really weird, its not at all related to this change, and the crasher there doesn't even reference Foundation. |
|
@swift-ci please test |
1 similar comment
|
@swift-ci please test |
|
Yup, ok. |
Windows x86_64 is a LLP64 environment.
unsigned longis notsize_t.Adjust the parameter to be the correct size for the allocator. This is
compatible with the previous implementation on LP64 environments.
Prefer to use the
__sync_bool_compare_and_swapover the MSVCintrinsic which is overly strong. While in the area, remove an unused
compare and swap method.