Re: How to make aot native-compilation and pdmp creation reproduceable

[Eli Zaretskii]

> From: Gerd Möllmann <gerd.moellmann@gmail.com>
> Cc: bjorn.bidar@thaodan.de,  emacs-devel@gnu.org
> Date: Mon, 09 Oct 2023 15:33:59 +0200
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> >> From: Gerd Möllmann <gerd.moellmann@gmail.com>
> >> Cc: bjorn.bidar@thaodan.de,  emacs-devel@gnu.org
> >> Date: Mon, 09 Oct 2023 13:55:30 +0200
> >> 
> >> Eli Zaretskii <eliz@gnu.org> writes:
> >> 
> >> >> My hypothesis is that x-win.eln changes between builds (print-circle
> >> >> etc.) => make-fingerprint computes a different ingerprint and so on.
> >> >
> >> > If x-win.el didn't change, x-win.eln will not be regenerated, so
> >> > something else is at work here.
> >> 
> >> But x-win.elc _did_ change.  To cite one of Björn's mails
> >
> > If x-win.elc or x-win.eln change between builds (meaning x-win.el or
> > one of the *.el files it 'require's changed), why do we expect the
> > builds to be identical?
> 
> AIAIU, the assumption that an .el file has changed doesn't hold.
> Instead, comp.el prints a constant list and produces a different $n=
> label in both cases.  The diff of the .eln's shows that.  If there are
> no further diffs in the .eln that weren't shown, everything else is
> identical.

Once again: why is x-win.el being ELN-compiled in the first place?
Are you saying that every time you say "make", we compile all the
preloaded *.el files, even if they didn't change?  That is definitely
not true.  So the question IMO is: why is x-win.eln being regenerated?