Re: Emacs 29.3 released

[Corwin Brust]

On Sun, Mar 24, 2024 at 1:46 PM Eli Zaretskii <eliz@gnu.org> wrote:
>
> > From: Corwin Brust <corwin@bru.st>
> > Date: Sun, 24 Mar 2024 12:52:21 -0500
> > Cc: emacs-devel@gnu.org
> > Can you see something I'm doing wrong, or might there be a problem
> > here which prevents with-native-compilation=aot from doing it's thing
> > for a build from a release tarball?
>
> I have no idea.  Comparing the configure script for Emacs 29.2 and
> Emacs 29.3 doesn't show any differences, and neither does comparing
> the various Makefile's.
>
> If you build Emacs 29.2 in this manner, do you get all the *.el files
> compiled into *.eln, or do you see the same problem with Emacs 29.2?

No, Building 29.2 from the release tarball (so, without running
autogen.sh) successfully precompiles all the ELNs as expected.

> If the latter, perhaps the reason has nothing to do with the new
> tarball, but with something else on your system?
>

I try to avoid making changes to my system between iterations of the
same version of Emacs (291. .. 29.3 are all built using the same
versions of things MSYS/MINGW, on my end)

> One note, though: if you say --with-native-compilation=aot, you don't
> need the NATIVE_FULL_AOT=1 part.  Not sure if this is relevant or not.
>

A good point.  I haven't tried removing this to see if anything
changes; 29.2 works properly, so I suspect this is a read herring but
I'll try removing it, nevertheless.

I do get all full AOT build if I add ./autogen.sh to the start of my
recipe -- which is really weird.  I find no differences between the
configure script provided with emacs-29.3 tarball and the one created
by running autogen.sh against the unpacked release tarball sources.  I
have no idea why AOT is working in the latter case but not the first.

I'm inclined to go ahead and publish using the version where I run
autogen.sh -- let me know if we prefer  that we track this issue down
first, instead.