On Sun, Feb 8, 2015 at 6:47 PM, Roberto De Ioris <[email protected]> wrote: > >> I don't have time to confirm right now, but the warning about scandir >> complains about the fourth argument, not the third, i.e. >> uwsgi_versionsort. >> >> I think it's due to const struct * v.s. const void *. >> > > > You mean scandir/versionsort declaration on your system is different from > this one ? > > http://man7.org/linux/man-pages/man3/scandirat.3.html
I dug into this a bit. Yes, glibc used to ship with a scandir() that did not have a POSIX-compliant decleration. See https://sourceware.org/git/?p=glibc.git;a=commit;h=eee6b14 and https://sourceware.org/bugzilla/show_bug.cgi?id=9759 This was fixed in glibc 2.10, CentOS 5 ships with 2.5. So you can revert 28dffb5, as NULL there is probably better than 0, or at least clearer. In any case, I don't care about CentOS 5, I just happened to be building on a build system that incorporates it. But in general this sort of thing will come up if you enable fatal warnings. I'm assuming you're doing that because you'd like more user bug reports, but rather than having to monkeypatch the build system to turn them off entirely it Would Be Nice if you could turn them off on a per-file basis or something via some feature :) _______________________________________________ uWSGI mailing list [email protected] http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
