On Tue, Oct 23, 2018 at 10:22 PM Stefan Behnel <stefan...@behnel.de> wrote: > > Robert Bradshaw schrieb am 22.10.2018 um 11:03: > > Given that https://bugs.python.org/issue4709 results in extension > > modules that seem to work, but silently produce completely incorrect > > answers, I'm thinking we should either disable our long-unpacking code > > on these platforms, or at the very least give a runtime error if we > > detect issues like sizeof(void*) != SIZEOF_VOID_P (or both). > > We could try to detect MinGW-64 and set the "MS_WIN64" macro before > importing the CPython headers. That seems to be the right fix. Can't say if > we also need to define the "MS_WIN32" macro, but would probably be better > to have it. CPython defines both for MSVC. > > There is a "__MINGW64__" macro to detect MinGW-64, it seems.
I created a minimal https://github.com/cython/cython/pull/2679 > > I'm not sure what has made this issue pop up more lately (Python 3), > > My guess is that MinGW is just fairly rarely used to build CPython > extensions overall. > > > but perhaps we should even consider releasing bugfixes for previous > > Cython versions. > > You mean a 0.28.6, or even older than that? I wish there were a good way to see what users are actually using in the wild... > As I said, MinGW does not seem to be used very widely… Yes, but it is used, and this is (IMHO) a really severe bug. _______________________________________________ cython-devel mailing list cython-devel@python.org https://mail.python.org/mailman/listinfo/cython-devel
