[fast-fixincludes] Several bugs and problems

[Robert Lipe]

> Looking into the fast-fixincludes failures on Digital UNIX and trying to
> bootstrap egcs on Ultrix V4.3, I found a couple of bugs and problems, some
> of which are fixed by the patch below:

Wow.   What a list.   Thanx!

> ** The !NO_BOGOSITY_LIMITS code claims to work around systems which limit
>    the number of fork()'s.  In my understanding, the limits that exist
>    (like Digital UNIX and Ultrix above) constrain the number of processes
>    *per user*, not per process, so there's nothing to be gained by forking
>    from a child process.

Earlier conversations with Bruce had led us to conclude that code should
instead just be banished.  It was there to work around a problem that
didn't really exist.

Bruce had a preliminary patch to pull this code out.  How do we feel about
yanking it at this point in the release cycle?

> ** Is there any reason not to use xmalloc() instead of checking malloc()
>    return values in many places?

Probably just history.   fixincl was a standalone program before it was
fitted into the autoconf/libiberty way of life.


RJL