On Samstag, 13. September 2025 20:21:34 CEST Pádraig Brady wrote: > On 13/09/2025 05:42, Bruno Haible wrote: > > Collin Funk wrote: > >> It would be nice to add a test module, but I'm conflicted on whether it > >> is too much of a chore. Since there isn't an easy way to check if a > >> hwcap is supported at runtime. > > > > A unit test for the hwcap_allowed function would be useful. That's a > > string parsing function, that apparently needs to deal with multiple > > cases: > > - GLIBC_TUNABLES not set, > > - GLIBC_TUNABLES set to empty, > > - other tunables than glibc.cpu.hwcaps, > > - glibc.cpu.hwcaps being the first one, > > - glibc.cpu.hwcaps being the last one, > > - glibc.cpu.hwcaps present but preceded and followed by other tunables, > > - glibc.cpu.hwcaps specifying multiple features. > > It's complicated enough to confuse me during code review... > > Thanks for the reviews all. > I'll push the attached later (with a ChangeLog entry). > > cheers, > Padraig
> + use other interfaces like getaxuval() instead. */ Typo: getaxuval → getauxval modules/cpu-supports-tests needs to depend on modules setenv unsetenv since these functions are missing on some platforms. > +#include <stdlib.h> > + > +/* Specification. */ > +#include "cpu-supports.h" It's better to #include "cpu-supports.h" directly after config.h. This verifies that it is a self-container header (i.e. does not accidentally use types or functions from <stdlib.h>). Bruno
