Building collect2 for C?

[Zack Weinberg]

Joe Buck <Joe.Buck@synopsys.COM> writes:

> DJ Delorie <dj@redhat.com> writes:
>> > It was based on libiberty.  DJGPP was just a sample of why.  IIRC, the
>> > patch created an API that wasn't platform-independent, defeating the
>> > purpose of putting it in libiberty in the first place.
>
> On Tue, Mar 16, 2004 at 11:03:16AM -0800, Zack Weinberg wrote:
>> It was only platform-dependent because you insist on using an
>> obsolete platform.  I am tempted to put DJGPP and MPW on the
>> obsoletions list for 3.5 just to eliminate this nonsense objection.
>
> If you want to mark DJGPP as an obsolete platform, you can propose it, and
> if 3/4 of the SC agrees, then and only then does it happen.
> Alternatively, if no SC member objects, we effectively have unanimous
> consent.  But no list that you maintain is in any way binding on its own,
> even though the SC in the past has blessed a list that you provided.

I did not intend to suggest otherwise.  I was being excessively
cranky.

However, I do think that DJ's objections to my patch are unfounded -
the API *could* be implemented, with restrictions, on platforms that
do not support pipes.  Furthermore, it doesn't break anything that
wasn't already broken.  Right now, DJGPP has exactly the same problem
with collect2 that Mingwin has.  In fact, my patch ought to make it
easier to support collect2 under DJGPP.  I don't think it should be
necessary to implement an intrinsically host-dependent feature on all
supported hosts in order to get the patch accepted.

I hope Ian's intended patch renders this whole discussion moot.

zw