在 2022/6/30 00:24, Ozkan Sezer 写道:
Lazy solution for me would be something like the following:
Does it look correct?
#if defined(__MINGW32__) && defined(__x86_64__)
#undef NtCurrentTeb
#define NtCurrentTeb MI_NtCurrentTeb
struct _TEB * MI_NtCurrentTeb(void) {
void *ret;
__asm__ __volatile
On 6/29/22, LIU Hao wrote:
> The compiler sees
>
> := *(unsigned long*) 42
>
> which itself is indeed dereferencing a strange pointer. Probably GCC should
> warn about only pointers
> to known data.
Lazy solution for me would be something like the following:
Does it look correct?
#if define
> The compiler sees
>
> := *(unsigned long*) 42
>
> which itself is indeed dereferencing a strange pointer. Probably GCC should
> warn about only pointers
> to known data.
Lazy solution for me would be something like the following:
Does it look correct?
#if defined(__MINGW32__) && defined(__
在 2022-06-29 22:48, Ozkan Sezer 写道:
Well, I'm not proficient with gcc inline asm.
So, whatever you guys think is best.
Also: is this not actually a gcc bug? Are they aware of it?
I don't think they are aware of it. Please file a report if you think it is
necessary.
The compiler sees
On 6/29/22, Ozkan Sezer wrote:
> On 6/29/22, LIU Hao wrote:
>> I have been aware of this too. The asm statement is passed an indirection
>> by
>> a constant offset in
>> the FS/GS segment, which itself is valid, but GCC thinks it was about the
>> default DS segment and
>> fails to calculate its s
On 6/29/22, LIU Hao wrote:
> I have been aware of this too. The asm statement is passed an indirection by
> a constant offset in
> the FS/GS segment, which itself is valid, but GCC thinks it was about the
> default DS segment and
> fails to calculate its size.
>
>
> This may be solved for GCC by c
在 2022-06-29 19:51, Ozkan Sezer 写道:
Noticed this in github CI runs of vkQuake. Is the warning bogus?
D:/a/_temp/msys64/mingw32/include/psdk_inc/intrin-impl.h:2164:1:
warning: array subscript 0 is outside array bounds of 'long unsigned
int[0]' [-Warray-bounds]
2164 | __buildreadseg(__readfsdwor
Noticed this in github CI runs of vkQuake. Is the warning bogus?
In function '__readfsdword',
inlined from 'NtCurrentTeb' at
D:/a/_temp/msys64/mingw32/include/winnt.h:2438:27,
inlined from '_mi_thread_id' at mimalloc/mimalloc-internal.h:814:21,
inlined from '_mi_is_main_thread' at mima
