sssd_flush_cache() errors ignored

[alejandro-colomar]

shadow/lib/sssd.c

Line 41 in 40bb89b

return -1;

shadow/lib/sssd.h

Line 13 in 40bb89b

#define sssd_flush_cache(service) (0)

This function (actually the macro when it's not being compiled) triggered an error while enabling -Werror=all, due to (all) callers ignoring the return value. Do we need to care about that return value when this function fails?

Cc: @jhrozek

[hallyn]

This is done after all other changes have been made, so trying to unwind the changes may resul

[alejandro-colomar]

This is done after all other changes have been made, so trying to unwind the changes may resul

?

[hallyn]

Egads, i wrote a lot more than that, that's kind of annoying.

tl;dr was that I don't think we can, if ssd_flush_cache() fails, undo the rest of the changes and return an error, because undoing the changes may also break. So we should syslog an error message.

Except that the error case you pointed to is where malloc(4) failed. In that case I don't expect syslog() to succeed either.

[alejandro-colomar]

Hmmm, okay, so all other error cases already do syslog, and this one it's pointless, so ignoring the return value is actually fine.

I guess it's okay. I'll then try to workaround the diagnostic.

[ikerexxe]

sssd deprecated the files-provider functionality in sssd-2.9.0, so we will probably remove this part of the code when all distributions have caught up.

[alejandro-colomar]

sssd deprecated the files-provider functionality in sssd-2.9.0, so we will probably remove this part of the code when all distributions have caught up.

Nice! Do you have an estimate of how distros are catching up with that?

[ikerexxe]

Nice! Do you have an estimate of how distros are catching up with that?

No, I'm not following this topic closely. The sssd version was released a year a go, so I'd probably give it another 1-2 years.

[alejandro-colomar]

Thanks!