Re: [RFC] Some sr.ht CI tweaking

"Kevin J. McCarthy" <[email protected]> Thu, 12 Feb 2026 11:46:00 +0800
Newsgroups	gmane.mail.mutt.devel
Message-ID	<aY1M-NGiYFxz9NLy@qinghai>
On Wed, Feb 11, 2026 at 06:21:12PM +0100, Rene Kita wrote:
>Inspired by commit 8b10799c (Fix iswblank() compilation error when
>configured --without-wc-funcs., 2026-01-26) I decided to take a look at
>the CI builds run on sr.ht.

Thanks Rene.

>My two main ideas were to use different build configurations and to
>compile with -Werror to make any warning abort the CI. I dropped the
>configure and build stage and added stages for each configuration which
>contain the configure and the build step. The inlined patches below are
>merely meant to get some feedback or start a discussion.

Sounds fine to me.  I agree with Alex.  Just drop the -pedantic.  Seeing
the workarounds makes me cringe.

>Some thoughts:
>- I don't think having a built run on Debian Stable is really
>  meaningful. Debian stable is most of the time quite old and only
>  selected patches will be backported. I would suggest to pick something
>  more bleeding edge like Arch.

As long as we have Ubuntu on gitlab I'm okay with this.  The idea is to
get a wider base of testing.

>Kevin, feel free to reject, a simple 'no' is enough. :-)
>
>commit b317691396a4f5a2505ccc095e17135f2bbd9081
>Author: Rene Kita <[email protected]>
>Date:   2026-02-04T09:12:18+01:00
>
>    Remove unused variable

OK.  I saw the warning too when testing --without-wc-funcs.

>commit 31f3299b356e4bdbfad2d9f2eb447c908c2869d9
>Author: Rene Kita <[email protected]>
>Date:   2026-02-04T07:07:34+01:00
>
>    Silence warning
>
>diff --git a/mutt_ssl.c b/mutt_ssl.c
>index ee08f943..762377e1 100644
>--- a/mutt_ssl.c
>+++ b/mutt_ssl.c
>@@ -708,6 +708,7 @@ static void ssl_err (sslsockdata *data, int err)
>       errmsg = "unknown error";
>   }
>
>+  (void)errmsg;  /* Silence warning when configured with --disable-debug */

OK.  Better than hacking the function to pieces with #ifdef DEBUG 
everwhere.

>commit 9b4284ade52b786e665c2daf8a784588a580a127
>Author: Rene Kita <[email protected]>
>Date:   2026-02-04T10:02:29+01:00
>
>    Fix build error for --without-wc-funcs on FreeBSD
>
>    mbyte.c:417:5: error: redefinition of '__wcwidth'
>    417 | int wcwidth (wchar_t wc)

Let me take another look at this.  I haven't studied this area much, so 
I'd like to understand how mutt even compiles if wchar.h is missing.

I wish the folks who had written all this stuff were still active... :-(

>commit 3a9124ac97f3f74d66c25be77f86d662854b6f76
>Author: Rene Kita <[email protected]>
>Date:   2026-02-06T11:39:11+01:00
>
>    Add sr.ht build manifest for OpenBSD

OK.

>commit 02cdc4d6c1e3786bd28a5a0245fe9772d4de31f7
>Author: Rene Kita <[email protected]>
>Date:   2026-02-03T18:11:43+01:00
>
>    Drop -Wno-long-long from CFLAGS
>
>    We already require C99.

OK.

>commit 8385ea5af95a2bf165147f8c459c3a8d7d90e0a7
>Author: Rene Kita <[email protected]>
>Date:   2026-02-03T18:15:22+01:00
>
>    Do not limit use of warnings to GCC
>
>    These are not compiler specific warnings, use them everywhere.

OK.

-- 
Kevin J. McCarthy
GPG Fingerprint: 8975 A9B3 3AA3 7910 385C  5308 ADEF 7684 8031 6BDA
signature.asc (application/pgp-signature, 833 B)