Matthew Parkinson

Results 160 comments of Matthew Parkinson

> ping :) Sorry, I forgot about this one.

Do you have a debug stack trace of when the flag is updated? The easiest way to debug this kind of crash is with time travel debugging with "WinDBG Preview",...

But `release` changes the `in_use` flag as well. So that would be useful to see. Also, are the stack traces being printed interleaved? The following looks strange. `VirtualAlloc` should not...

It is not calling `inner_release` in the destructor of ```C++ static thread_local OnDestruct tidier; ``` This works on other implementations of the CRT. I guess it could be a mingw...

I don't really want to support MinGW as I don't see much demand, and this makes me even less keen. It will lead to memory leaks on MinGW. Your mini...

>Sadly, the bug appears on GCC9, GCC10 and clang-10. I assume this is with MinGW?

Three year old issues. I am not particularly hopeful it will be fixed.

@mati865 and @sbabbi thanks. I think our issue is slightly different, it `__cxa_thread_atexit` does not appear to be doing the right thing with regard to the TLS implementation. The things...

The `ctz` issues for MinGW was also affecting `clang-cl`, so I have done the ctz fix suggested by @davidchisnall in #242.

@mati865 > there is no bug in mingw-w64 itself but within the compiler used together with it. Is there a combination we should recommend for building `snmalloc` then?