Re: emacs-28 9c222b9: Port to C compilers that lack size-0 arrays

[Eli Zaretskii]

> Date: Fri, 3 Dec 2021 11:42:08 -0800
> Cc: emacs-devel@gnu.org
> From: Paul Eggert <eggert@cs.ucla.edu>
> 
> On 12/3/21 10:29, Eli Zaretskii wrote:
> >> What I experienced was a ton of diagnostics with both the Oracle Solaris
> >> and the IBM AIX C compilers, which made it difficult for me to see the
> >> real problems on those platforms (of which there were some).
> > Any reason why we should be concerned about these two platforms?
> 
> Emacs has run on Solaris and AIX for decades, these platforms are stable 
> and are supported by Oracle and IBM respectively, and there's very 
> little trouble keeping Emacs working on these platforms.

Emacs still did compile on those platforms before your changes.  So we
are not talking about dropping support for them or breaking the build
on them, we are talking about avoiding warnings.

> And it's not just these two platforms. My bigger motivation for this 
> particular change was that the old code violates a constraint specified 
> by the C standard. Even though the old code works on GCC, Clang and 
> MSVC, there's a reasonable chance it won't work on less-commonly-used 
> compilers and there's little point to taking that risk when the fix is 
> so simple.

That's okay, but such cleanup changes should be done on master, not on
the release branch during a pretest.  The time for cleanups on the
release branch has come and gone.  (And we don't support MSVC builds
for a very long time now.)

So I'm inclined to revert that changeset on the release branch.