On Tue, 19 Nov 2002, Alistair Crooks wrote:

> On Fri, Nov 15, 2002 at 11:02:29AM -0800, Jeremy C. Reed wrote:
> > Why do some packages check if software is already in the base (default)
> > installs?
> >
> > For example, sysutils/user doesn't want to install if it isn't NetBSD and
> > /usr/sbin/user exists.
> >
> > This doesn't matter.
> >
> > Does it matter if you choose to install both /usr/pkg/sbin/useradd and
> > /usr/sbin/useradd?
>
> It matters to me, since I dislike playing Russian roulette with PATH
> settings - I have always found that these things come back and bite me
> in the bottom.

...

> emulation.  That was when the DEPEND stuff grew an indication of the
> full pathname of the binary that was needed.

Thanks for your explanation. I agree that full pathname can be helpful.

Could we add a setting that will by-pass these checks?

That way someone could set that variable to allow installing filenames of
same name without having to manually go through and patch various files.

Again, note that we already have packages that install files with same
names, such as CUPS' lpr, lpq, lprm, lpc, lpstat (under some OSes).

Also, several other executables have same names on some operating systems
and (I don't think are checked) in our pkgsrc, such as: addr2line, ar, as,
bunzip2, bzcat, bzip2, bzip2recover, c++filt, cvs, cvsbug, gettext, gprof,
ld, lp, lpq, lpr, lprm, msgcmp, msgcomm, msgfmt, msgmerge, msgunfmt, nm,
objcopy, objdump, openssl, perl, ranlib, rcs2log, readelf, sftp, size,
strings, strip, and xgettext.

   Jeremy C. Reed
   http://bsd.reedmedia.net/