Re: Abysmal state of GTK build

[Eli Zaretskii]

> Date: Wed, 24 Aug 2022 08:38:38 +0200
> Cc: emacs-devel@gnu.org, luangruo@yahoo.com
> From: Gerd Möllmann <gerd.moellmann@gmail.com>
> 
> On 22-08-23 18:06 , Eli Zaretskii wrote:
> >> From: Lars Ingebrigtsen <larsi@gnus.org>
> >> Cc: eliz@gnu.org,  emacs-devel@gnu.org,  luangruo@yahoo.com
> >> Date: Tue, 23 Aug 2022 16:43:40 +0200
> >>
> >> The most recalcitrant thing is ./configure, of course -- it's
> >> single-threaded.
> > Are you using -C ?  It makes the configure script fly.  Most of the
> > changes in the script nowadays don't need a full reconfiguration, and
> > seldom even add new variables.
> >
> Interesting.   How are you using --config-cache?

I just make a point of using it the first time I build a new
worktree.  Thereafter, I just use "make", and if it needs to re-run
configure, it uses -C automatically.

> I tried here two './configure --config-cache --with-native-compilation' 
> in a row (no config.cache initially), and the second run failed because 
> of something I don't understand.  Basically, it first says libgccjit.h 
> found, and then fails because it can't find libgccjit.h.
> 
> Is this a bug, or am I doing it wrong?

It's some kind of bug, but I'm not sure where.  I have this from time
to time on one particular system where I regularly build Emacs, but
not on others.  Usually, a second "make" right after that succeeds.

I always thought that it's because some system-wide snafu unrelated to
Emacs on that single system.  But now I'm not so sure.

(There's no longer any need to run 'configure', except, perhaps, the
first time you build a clean clone.  Just "make" does everything,
including running autogen.sh and the configure script, if needed.)