On Sat, 9 Mar 2002, matthew green wrote:

:    "Because it doesn't do anything useful for you" doesn't mean "throw it out"
:    -- some other people may find utility in it.  I've needed COMPAT_AOUT a
:    total of 6 times already for migration purposes.

: perhaps if uucp and rcs were broken and unmaintained this argument
: would be valid.

They're certainly unmaintained, or at least RCS is.  Besides, we don't use
RCS in any of our source tree; CVS is there.  RCS should be ejected with
extreme prejudice (to catch a phrase)....

: the fact is that COMPAT_AOUT f*cks you over and it should be ejected with
: extreme prejudice.

In its current form, it has problems with some things, but only if it's
enabled.  It doesn't "f*ck you over", or at least, it doesn't mess you up if
you know that it's part of the NetBSD emulation system (see below).

: anyone running a system with a.out binaries that they *USE* really
: doesn't want to have COMPAT_AOUT.  why?  say your shell is aout..
: you type "cd /usr" in it.  where do you end up?  "/emul/aout/usr".

The same thing happens with Linux, SVR4, and all others.  I've given
solutions to this problem in the past (>2 years ago, when I didn't have the
knowledge on how to fix it properly).  I may revisit that.

However, this is not an artifact of only COMPAT_AOUT, so don't try to paint
it as such.

=====

Note that if we do centralize the pathname translation, every one of these
bits becomes moot; if someone wants a.out (or even native ELF, as was noted
elsewhere on the thread!) to have translation, it becomes a one-line source
diff.

-- 
-- Todd Vierling <tv@wasabisystems.com>  *  Wasabi & NetBSD:  Run with it.
-- CDs, Integration, Embedding, Support -- http://www.wasabisystems.com/