This is the mail archive of the
libc-hacker@sourceware.cygnus.com
mailing list for the glibc project.
Re: Alpha failure [was: Re: egcs test failures under glibc 2.0.98]
- To: Philip Blundell <philb@gnu.org>
- Subject: Re: Alpha failure [was: Re: egcs test failures under glibc 2.0.98]
- From: Andreas Schwab <schwab@issan.informatik.uni-dortmund.de>
- Date: 26 Oct 1998 10:53:57 +0100
- Cc: hjl@lucon.org (H.J. Lu), aj@arthur.rhein-neckar.de (Andreas Jaeger), libc-alpha@cygnus.com
- References: <E0zXVGo-00036Z-00@kings-cross.london.uk.eu.org>
Philip Blundell <philb@gnu.org> writes:
|> >I was trying to build glibc in CVS as of Oct 24 07:00am PDT on
|> >alpha. But egcs 1.1 used so much memory on one file. It ran out
|> >of memory. Do I need a new egcs to compile it on alpha. BTW, I
|> >am using the kernel header files from Linux 2.1.126.
|>
|> I had the same problem (with nss/nsswitch.c). The compiler used up over 100MB
|> of virtual memory and eventually died with SIGSEGV. I compiled that file by
|> hand without -O and there was no problem.
This is a feature :-) due to the huge macros in bits/string2.h. And it is
also a consequence of the improvements of __builtin_constant_p in egcs,
which causes it to survive the tree->rtl conversion until cse, leaving
behind a big pile of rtl garbage when it is finally replaced by a
constant.
Andreas.
--
Andreas Schwab "And now for something
schwab@issan.cs.uni-dortmund.de completely different"
schwab@gnu.org