Thiemo Seufer wrote:
> Camm Maguire wrote:
> > Greetings! I've recently put in native object code relocation on mips
> > for gclcvs following the lush example, and tested on an Indy locally
> > (which has since died.) mipsel autobuilds just fine, while mips give
> > a bus error right at the end. Could this possibly be related to some
> > physicall error on the machine?
> Yes, this is quite possible.
> > Is there a way in which code which
> > makes no reference to byte order might logically fail on only one of
> > mips/mipsel? Lastly, if the fault is with gclcvs and this needs
> > debugging, can anyone arrange ssh access to a remote mips machine?
> Hm, I'll have a look.
It built with the appended patch. I appears the mipsel version succeded
to build because it didn't select native object relocation (the mips
pattern matched first). Generally, canonical GNU host ids should be
checked with at least a cpu-vendor-os pattern to avoid mismatches, that
is *-*-* as the minimum catch-all.
(What I used below matches also for e.g. mipsisa32el-unknown-linux-uclibc.)
diff -urpN gclcvs-2.7.0.orig/configure.in gclcvs-2.7.0/configure.in
--- gclcvs-2.7.0.orig/configure.in 2005-04-14 23:06:33.000000000 +0200
+++ gclcvs-2.7.0/configure.in 2005-04-19 15:10:43.000000000 +0200
@@ -109,12 +109,12 @@ case $canonical in
diff -urpN gclcvs-2.7.0.orig/o/sfaslbfd_mips.c gclcvs-2.7.0/o/sfaslbfd_mips.c
--- gclcvs-2.7.0.orig/o/sfaslbfd_mips.c 2005-04-14 23:19:59.000000000 +0200
+++ gclcvs-2.7.0/o/sfaslbfd_mips.c 2005-04-19 05:32:32.000000000 +0200
@@ -471,7 +471,7 @@ mipself_fix_relocs(bfd *b,asymbol ** sym
gpdisp = symbols[i];
gpdisp->value = 0;
gpdisp->flags = BSF_LOCAL|BSF_SECTION_SYM;
- gpdisp->section = info->sgot ? info->sgot : info->sgot : bfd_abs_section_ptr;
+ gpdisp->section = info->sgot ? info->sgot : bfd_abs_section_ptr;
gpdisp->udata.p = NULL;
- From: Camm Maguire <email@example.com>
- Re: gclcvs
- From: Thiemo Seufer <firstname.lastname@example.org>