GNU bug report logs - #58254
29.0.50; Untracked lib/assert.h with Clang 14

Previous Next

Full log

View this message in rfc822 format

From: "Basil L. Contovounesios" <contovob <at> tcd.ie>
To: Paul Eggert <eggert <at> cs.ucla.edu>
Cc: 58254 <at> debbugs.gnu.org
Subject: bug#58254: 29.0.50; Untracked lib/assert.h with Clang 14
Date: Mon, 03 Oct 2022 16:00:51 +0300

[Message part 1 (text/plain, inline)]

Paul Eggert [2022-10-02 13:20 -0700] wrote:

> On 10/2/22 09:10, Basil L. Contovounesios wrote:
>> This file is not generated with the default CC=gcc; there is only a
>> lib/assert.in.h present.
>
> That's odd. As far as I can see, lib/assert.h is needed only if you need to
> compile lib/filevercmp.c, lib/nanosleep.c, lib/pipe2.c, lib/rawmemchr.c, or
> lib/strtoimax.c, and I don't see why choosing clang over gcc would affect
> whether these files need to be compiled.

Out of these, only rawmemchr.c and strtoimax.c are not compiled.
The rest are compiled with both Clang and GCC.

I suspect there's a different culprit: skimming through the attached
build output and config.log files, one set with Clang and the other with
GCC, I see that, in the GCC case, static_assert is determined to be an
<assert.h> macro, whereas in the Clang case the static_assert test
fails.  As a result, gl_ASSERT_H sets GL_GENERATE_ASSERT_H=true.
Is that correct?

Thanks,

-- 
Basil

[emacs-bug58254.tar.gz (application/gzip, attachment)]

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.