Re: Guile 64-bit Windows support, redux

Jonas Hahnfeld via "Developers list for Guile, the GNU extensibility library" <[email protected]> Fri, 28 Feb 2025 22:28:46 +0100

Newsgroups	gmane.lisp.guile.devel
Message-ID	<[email protected]>

On Tue, 2025-02-25 at 23:07 +0000, Mike Gran wrote:
> First off, these patches look good to me. They are necessary if we want to
> try to support Windows. Plus Jonas figured out a couple of the errors I never managed
> to work out with Lightning and the VM. So I'm very impressed.
> 
> A couple of questions...
> 
> From what I can tell, these patches are from a cross-compilation toolchain.
> Is that correct?

Yes, for LilyPond we build Guile in a MinGW corss-compilation setup.

> They won't build from source with a native MinGW toolchain because
> of the DLL version number problem and a couple of other tweaks.
> Do we care about making them build with a native toolchain?
> 
> The DLL number version problem is that the native MinGW tools build DLLs with
> a major version number:  libfoo.so.3.2.1 => libfoo-3.dll
> 
> If we do care about building with the native MinGW toolchain,
> in my humble opinion, there isn't much value in supporting then many
> versions of MinGW other than UCRT64. Does LilyPond care
> about the old MinGWs? MINGW32? MINGW64? CLANG64?

At the moment, we use the x86_64-w64-mingw32-* prefixed tools. But if I
understand https://www.msys2.org/docs/environments/ correctly, the
UCRT64 environment is only relevant for native compilation? Also not
really an expert in native building, I stopped when cross-compilation
worked and I got my 64-bit binaries out...

signature.asc (application/pgp-signature, 228 B)