Re: Native compilation on Windows, was Re: Bootstrap Compilation Speed

[phillip . lord]

On 2022-01-15 19:06, H. Dieter Wilhelm wrote:
> Corwin Brust <corwin@bru.st> writes:
> >  > phillip.lord@russet.org.uk writes:
> >  The performance that I got on Windows builds when I did do it was 
> > highly
> >  variable (between half an hour and three), but I think this was 
> > mostly
> >  the shared CPU infrastructure I was using. I never worked out how to 
> > get
> >  native comp working in a way that I could package it for windows, so 
> > the
> >  builds were all without it.
> >
> > This is my present situation also.  I can build with native comp but
> > not (yet) found a way to package that version.
>
> I don't understand yet the packaging requirements, is it not possible 
> to
> copy additionally the native-lisp/ folder to the package?
>
> So far I could compile (locally) the current pretest 28.0.91 version
> --with-native-compilation.
>
> Then I tried to build (with build.zips.sh) a snapshot of the
> masterbranch --with-native-compilation on the same system.  And now
> ./configure is complaining:
>
>   configure: error: Elisp native compiler was requested, but libgccjit
> was not found.
>
> Does Emacs-29 has different requirements?


The problem is that native comp needs libgccjit at compile time but also 
at
runtime. To make an executable install package like we have at the 
moment
essentially means packaging libgccjit. Which means gcc and half of the 
msys2
toolchain.

We were there a long time back when Emacs used to be packaged with lots 
of other
stuff, including a full python installation. It didn't make sense then 
and it's
not a great thing to go back to.

So, I think, Emacs releases for windows will be without native-comp. 
Anyone who
wants it will have to install it over an msys2 installation, or with 
chocolaty
or something equivalent.

Phil