On Tuesday, 6 November 2018 at 15:08:47 UTC, Guillaume Piolat
wrote:
On Monday, 5 November 2018 at 14:46:25 UTC, kinke wrote:
On Monday, 5 November 2018 at 13:53:50 UTC, Guillaume Piolat
wrote:
Very cool! And very scary too, we'll have to verify the
transcendantal precision and memcpy performance :)
Don't worry, we're not using anything from the MinGW runtime
itself - just functions exported by the MS DLLs and some
minimal startup/shutdown skeleton (`mainCRTStartup` entrypoint
etc.). We (incl. DMD) are just using their .def files as basis
for the generated import libraries.
I don't understand, llvm_exp for example translate to a call to
the C stdlib exp() IIRC. That is currently in the MS runtime,
no?
Yep, and still is, as I said. We don't use any MinGW functions at
all, so nothing changes in this regard, no need to worry about
MinGW's 80-bit `long double`, their different C++ mangling, their
.a static library format or their DWARF debug info. You're still
targeting windows-msvc. We only need the MinGW .def files because
Microsoft shamefully doesn't allow distribution of their static &
import libs.
As stated in the readme, using the MS toolchain is obviously
still supported.
