Dear Maintainer,

If I'm not mistaken, there's now an issue with magit 4.3.0 running Emacs 30.1 in testing.

Issueing M-x magit-status after Emacs has started, from a directory containing a git checkout of a repo, I get an error about magit-commit-absorb

The Messages buffer reports :
require: Symbol’s value as variable is void: magit-commit-absorb

The *Async-native-compile-log* buffer reports:

Compiling /usr/share/emacs/site-lisp/elpa/magit-4.3.0/magit-sequence.el...
/usr/share/emacs/site-lisp/elpa/magit-4.3.0/magit-sequence.el: Error Symbol's value as variable is void: magit-commit-absorb
Compiling /usr/share/emacs/site-lisp/elpa/magit-4.3.0/magit-notes.el...
/usr/share/emacs/site-lisp/elpa/magit-4.3.0/magit-notes.el: Error Symbol's value as variable is void: magit-commit-absorb
Compiling /usr/share/emacs/site-lisp/elpa/magit-4.3.0/magit-worktree.el...
/usr/share/emacs/site-lisp/elpa/magit-4.3.0/magit-worktree.el: Error Symbol's value as variable is void: magit-commit-absorb
Compiling /usr/share/emacs/site-lisp/elpa/magit-4.3.0/magit-tag.el...
/usr/share/emacs/site-lisp/elpa/magit-4.3.0/magit-tag.el: Error Symbol's value as variable is void: magit-commit-absorb
Compiling /usr/share/emacs/site-lisp/elpa/magit-4.3.0/magit-merge.el...
/usr/share/emacs/site-lisp/elpa/magit-4.3.0/magit-merge.el: Error Symbol's value as variable is void: magit-commit-absorb
Compiling /usr/share/emacs/site-lisp/elpa/magit-4.3.0/magit-branch.el...
/usr/share/emacs/site-lisp/elpa/magit-4.3.0/magit-branch.el: Error Symbol's value as variable is void: magit-commit-absorb
Compiling /usr/share/emacs/site-lisp/elpa/magit-4.3.0/magit-reset.el...
/usr/share/emacs/site-lisp/elpa/magit-4.3.0/magit-reset.el: Error Symbol's value as variable is void: magit-commit-absorb
Compiling /usr/share/emacs/site-lisp/elpa/magit-4.3.0/magit-files.el...
/usr/share/emacs/site-lisp/elpa/magit-4.3.0/magit-files.el: Error Symbol's value as variable is void: magit-commit-absorb
Compiling /usr/share/emacs/site-lisp/elpa/magit-4.3.0/magit-refs.el...
/usr/share/emacs/site-lisp/elpa/magit-4.3.0/magit-refs.el: Error Symbol's value as variable is void: magit-commit-absorb
Compilation finished.

Hope this is indeed linked to the Debian package, and not some strange customization on my side.

Best regards,

Olivier Berger <olivier.berger@telecom-sudparis.eu> writes:
Emacs and "M-x magit-status" on a git directory, and my
*Async-native-compile-log* doesn't show any warning/error regarding
magit.

Can you try starting a clean Emacs using "emacs -Q", load only magit
(probably also transient and other dependencies), and try to reproduce
this?

Olivier Berger <olivier.berger@telecom-sudparis.eu> writes:
Emacs and "M-x magit-status" on a git directory, and my
*Async-native-compile-log* doesn't show any warning/error regarding
magit.

Can you try starting a clean Emacs using "emacs -Q", load only magit
(probably also transient and other dependencies), and try to reproduce
this?

Emacs 30.1 in testing.
containing a git checkout of a repo, I get an error about magit-commit-
absorb

I also experience this issue.

It seems that reinstall the elpa-magit package or deleting native-
compiled files under ~/.config/emacs/eln-cache helps, because the error
disappears just after that, but it comes back later after Emacs has
been restarted.

Also I manage to reproduce the problem by running “emacs -q” and then
just “(require 'magit)”.

Le mercredi 12 mars 2025 à 16:32 +0100, Sébastien Villemot a écrit :

I think I found the cause of the problem. There was an old bytecode-
compiled version of transient under /usr/share/emacs/site-
lisp/elpa/transient-0.3.7/

I removed that directory, then also removed ~/.config/emacs/eln-cache/,
and everything seems to be working fine again.

So ultimately this may be a bug in elpa-transient. It would be useful
to test a clean upgrade from bookworm to trixie to see whether this is
reproducible.

Sébastien Villemot <sebastien@debian.org> writes:

This sounds like the upgrade issue that Rob, David, and Sean has been
trying to address in dh-elpa 2.1.0 and later.  transient 0.3.7 is from
Bookworm, so I think this could probably happen when transient was
upgraded with dh-elpa <2.1.0.

Hello,

Yeah, it's probably this one.