Re: gcc bootstrap on 3.1 (probably 3.0.4) on i586-pc-sco3.2v5.0.2 --with-gnu-as

[Robert Lipe <robertlipe at usa dot net>]

rbrown64@csc.com.au wrote:

> Now that COFF (support in gcc 3 on sco3.2v5) sleeps with the fishes,
>  is sco3.2v5.0.x a sufficiently standard ELF
> platform to enable ld in binutils? It too is coming up to a release (2.12).

If GNU ld worked well enough to turn in good testresults on a gcc run,
I'd support turning it on, but...

> FAIL: bootstrap               # Undefined basename

This is 4/10 of the failures you describe.  I don't know what this
means.  There is a basename command in /bin.

> FAIL: bootstrap with --static       # Indigestion on /lib/libc.a - which is
> still COFF?

There's the rub and, I suspect, the dealbreaker.  OpenServer's libraries
are wierd.  I can't recall the details, but I seem to remember them
containing both ELF and COFF objects in the archives and the native
linker knowing how to sort that all out.

> FAIL: cdtest                  # tmpdir/cdtest: syntax error at line 1: `('
> unexpected: cannot execute binary file
> FAIL: cdtest with -Ur         # " "

This probably means it's emitting an incorrectly formed binary.

>                 === ld Summary ===
> 
> # of expected passes            15
> # of unexpected failures        10
> # of expected failures          1

Failing 2/3 of the tests sounds pretty serious to me.  Does it seem to
work "well enougn" in the real world?

So even though we finally gave up on GCC supporting both COFF and ELF
and that's probably enough to make GAS viable on this target, the linker
still gets exposed to system facilities that have both and I don't know
that it won't be blinded by that light.  Maybe it can be taught to cope;
maybe it already has been.

RJL