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

[Eli Zaretskii <eliz@gnu.org>]

> From: Corwin Brust <corwin@bru.st>
> Date: Thu, 3 Feb 2022 07:57:32 -0600
> Cc: "H. Dieter Wilhelm" <dieter@duenenhof-wilhelm.de>, Arash Esbati <arash@gnu.org>, 
> 	Emacs developers <emacs-devel@gnu.org>, Phillip Lord <phillip.lord@russet.org.uk>
> 
> > > As an interesting aside:  I see that it *is* loading the natively
> > > compiled sources that were build and packaged with Emacs.  This is
> > > despite getting a nil return from `native-comp-avialable-p', which
> > > suggests that even people without a local libgccjit+gcc (e.g. a
> > > working MSYS) can benefit from (the bundled) nativly compiled sources
> > > we distribute.  [[stroking chin myscheviously]]
> >
> > This is the expected behavior.  Only native-compilation of *.el files
> > for which no *.eln files are provided is disabled without libgccjit
> > and Binutils installation.  Loading of existing *.eln files doesn't
> > need libgccjit.
> >
> > So I'm not sure why you are surprised.
> 
> Now that ask, I'm not sure either.
> 
> Should we consider adding NATIVE_FULL_AOT when building the windows
> distribution?

You could, but that still means users cannot modify the *.el files and
get them recompiled.

Whether NATIVE_FULL_AOT is worth is: it is yet to be determined.  I
tend to think we should provide a default build, and only use
NATIVE_FULL_AOT if other platforms do that (which will not happen
before Emacs 28 is released and we collect feedback from users and
downstream distributions).