Re: [PATCH] fcntl.h should provide mode_t, off_t and pid_t

[Roland McGrath <roland at redhat dot com>]

> It isn't because on Linux configurations _all_ types are just always
> pulled by <bits/fcntl.h> from <sys/types.h>, as is also done in a bunch
> of other headers.

Ok.  I believe this is kosher in POSIX under the general "*_t" rule.
I don't really recall, but I think we just decided it was simpler to
use <sys/types.h> in various places.

> > If it is missing only in Hurd configurations, we should figure out how
> > the difference arises before changing anything.
> 
> That's the difference.

Might as well just change the Hurd's bits/*.h headers to match the Linux
ones in their <sys/types.h> use and not worry about it in detail.

> BTW, I've actually just ended up generating systematic tests, and we

We have the stuff in conform/ for that.  I guess it's not used by make
check or anything.  May it's bit-rotted.  Would you like to work on fixing
that up and making sure it will catch everything that you have observed?


Thanks,
Roland