Re: Emacs 27.0.90 is out!

[Paul Eggert <eggert@cs.ucla.edu>]

On 3/4/20 10:40 PM, Eli Zaretskii wrote:

> I'm not yet sure which part(s) of the changeset directly shut up the
> "-Walloc-size-larger-than=N" warnings.  Is that the part that adds the
> calls to 'assume'?

Yes.

> If so, we already call eassume in many places; it
> won't be a problem to add a few more, and fix the warnings on the
> emacs-27 branch.

I'd rather not backport this stuff into the emacs-27 branch, since the 
eassume calls do not fix bugs and can change the code generated and 
therefore can inadvertently introduce bugs. People building on emacs-27 
can either live with the warnings, or use ordinary GCC options instead 
of unusual ones, or not enable warnings.

Pacifying GCC when Emacs is built with unusual options is a low-priority 
thing to do, as we can't feasibly pacify GCC's false alarms under every 
option combination (there are far too many combinations and pacifying 
GCC in every instance would cause make-work maintenance and complicate 
Emacs and wouldn't make Emacs more reliable). Instead, we should merely 
suggest to people to not enable GCC warnings when they are employing 
unusual GCC options.

> And what about a similar warning I reported in bug#39857?

That's a false alarm, and I suppose it's a bug in the particular GCC 
build that you used. I expect the simplest workaround is stick to the 
usual compiler options (i.e., avoid -Og), or to disable GCC warnings. 
For what it's worth I can't reproduce the problem on Fedora 31 x86 
(i.e., 32-bit) with GCC 9.2.1 20190827 (Red Hat 9.2.1-1), so perhaps you 
can work around the problem by using a newer GCC.