Am Mon, 10 Aug 2015 12:36:32 +0200 schrieb "Iain Buclaw via D.gnu" <d.gnu@puremagic.com>:
> On 6 August 2015 at 13:19, Mike Parker via D.gnu <d.gnu@puremagic.com> > wrote: > > > On Thursday, 6 August 2015 at 11:07:33 UTC, Mike Parker wrote: > > > >> Haven't tried the 64-bit binaries yet, but any attempt to compile > >> an executable with the 32-bit binaries results in this: > >> > > > > OK, I've tried the x86_64-w64-mingw32 binaries. When compiling with > > -m32, I get the same errors as above. The default -m64 works. > > > > > Bug report? > Looks like it's some kind of problem with underscore prefixing. `D2rt5tlsgc14processGCMarksFNbPvMDFNbPvZiZv' should likely have an underscore appended. Our MinGW port is probably mostly broken, but we need a good way to run the testsuite for MinGW to really have a useful MinGW port. Once I've got some time for this I want to take the standard DMD test runner, make it a real stand-alone program and add support for cross compiler testing using ssh/scp. Once we have that it should be possible to reliably test the MinGW port. BTW: As MinGW mostly uses the standard MS runtime and upstream druntime/phobos now supports these we'll probably only have to update version blocks. TLS could be a problem, but we could just compile with emutls for now which should work.
