[6003] in bugtraq

home help back first fref pref prev next nref lref last post

Re: GCC 2.7.? /tmp files

daemon@ATHENA.MIT.EDU (Zack Weinberg)
Mon Jan 19 11:30:52 1998

Date: 	Sun, 18 Jan 1998 13:51:12 -0500
Reply-To: Zack Weinberg <zack@RABI.PHYS.COLUMBIA.EDU>
From: Zack Weinberg <zack@RABI.PHYS.COLUMBIA.EDU>
X-To:         dichro-bugtraq@rcpt.to
To: BUGTRAQ@NETSPACE.ORG
In-Reply-To:  Your message of "Sat, 17 Jan 1998 21:21:22 +0800." 
              <s2vg1mntewd.fsf@mermaid.ucc.gu.uwa.edu.au>

On Sat, 17 Jan 1998 21:21:22 +0800, dichro-bugtraq@RCPT.TO wrote:
>>>>>> "NB" == Niels Bakker <niels@EURO.NET> writes:
>
>    NB> Unless you specify -pipe.  Perhaps our kind
>    NB> totally-not-overworked BUGTRAQ moderator :-) would like to see
>    NB> whether GNU autoconf sets -pipe automatically when gcc is
>    NB> detected?  (I think so but unfortunately don't have much
>    NB> sources handy at the moment.)
>
>  Ah, but -pipe has it's own problems with autoconf. It stops various
>tests working - eg, the asm tests for ssh, and various assembler tests
>for GNU binutils (from memory). As a complete solution, this lacks a
>certain something :)

This is because of another bug in gcc 2.7.x: -pipe applied to a .s
file will hang the compiler.  The compiler driver incorrectly assumes
that when -pipe is in use, there's something ahead of the assembler in
the pipeline.

Applying this patch to /.../gcc-lib/MACHINE/VERSION/specs will fix it:

=cut here=
--- specs~      Wed Jun 26 15:14:20 1996
+++ specs       Sun Jan 18 13:46:30 1998
@@ -4,3 +4,3 @@
 *asm_final:
-%{pipe:-}
+%|

=cut here=

and if anyone cares, I have autoconf macros to detect the bug and
hack around it.

The bug is fixed in egcs and I hope also in 2.8.

zw

home help back first fref pref prev next nref lref last post