Re: [bug-gnulib] same, utimecmp should depend on minmax

[Paul Eggert]

Derek Price <address@hidden> writes:

> Isn't there some sense to assuming that if <limits.h> or <sys/param.h>
> define MIN & MAX on a system, then including said header is the
> "correct" way of getting the definition on one of these systems?

Perhaps, if the macros there are correct.  Sometimes a header will
define MIN and not MAX, or vice versa.

I just checked, and 5 headers on my Debian 3.0r1 host define MIN
directly: sys/param.h, linux/cycx_drv.h, linux/efs_fs.h,
linux/if_tun.h, and linux/isicom.h.  I don't know how many define
it indirectly.

Nine headers on my Solaris 10 box define MIN directly:

Mrm/Mrm.h
Xm/TextP.h
fm/fmd_api.h
fm/fmd_fmri.h
sys/fs/hsfs_rrip.h
sys/ldterm.h
sys/mdb_modapi.h
sys/sysmacros.h
utility.h

Again, I don't know how many define it indirectly.

More and more this is starting to sound like a loser situation.
Perhaps we should give up on these predecessor #includes entirely, and
warn users that they must include minmax.h after they include any file
that might possibly include a system file.