Hi Bruno, On Thu, Nov 14, 2024 at 04:14:26PM GMT, Bruno Haible wrote: > I promised: > > - see how this new warning behaves on the Gnulib source code
Thanks!
> The two good news are:
> - The warning is really helpful. With it, I could find all relevant uses of
> '0' quickly. Patch attached and committed.
> - There were no false positives.
Lovely! :-)
> However, there are warnings that are not in the responsibility of a package
> maintainer:
>
> * Anywhere a pthread_mutex_t is initialized, the warning appears. This is
> because on glibc, the macro __PTHREAD_MUTEX_INITIALIZER(__kind)
> expands to
> 0, 0, 0, 0, __kind, 0, 0, { 0, 0 }
> It should better expand to
> 0, 0, 0, 0, __kind, 0, 0, { NULL, NULL }
It would be interesting to send a patch to glibc for removing all uses
of 0 as a null pointer constant within glibc, including that one.
I can try this, unless someone beats me to it.
> * gperf generated code has these warnings:
> unicase/special-casing-table.gperf:171:10: warning: zero as null pointer
> constant [-Wzero-as-null-pointer-constant]
> unictype/categ_byname.gperf:115:10: warning: zero as null pointer
> constant [-Wzero-as-null-pointer-constant]
> unictype/incb_byname.gperf:60:10: warning: zero as null pointer constant
> [-Wzero-as-null-pointer-constant]
> unictype/joininggroup_byname.gperf:131:10: warning: zero as null pointer
> constant [-Wzero-as-null-pointer-constant]
> unictype/pr_byname.gperf:114:10: warning: zero as null pointer constant
> [-Wzero-as-null-pointer-constant]
> unictype/scripts_byname.gperf:160:10: warning: zero as null pointer
> constant [-Wzero-as-null-pointer-constant]
> uninorm/composition-table.gperf:995:10: warning: zero as null pointer
> constant [-Wzero-as-null-pointer-constant]
Same for gperf. Would you mind patching that one?
> These can only be fixed with a new glibc release and a new gperf release.
>
> So, as things stand now, we can recommend the warning option to package
> maintainers. But I am not in favour of letting Gnulib enable it by default,
> since it will invariably lead to warnings (in the two cases mentioned above)
> that the maintainer cannot silence. Only once there have been glibc + gperf
> releases and we have waited 2 years until most maintainers have a system with
> these newer releases, will it make sense to enable it by default.
Makes sense.
>
> Bruno
Have a lovely day!
Alex
--
<https://www.alejandro-colomar.es/>
signature.asc
Description: PGP signature
