Re: Mainline bootstrap failure (Re: libstdc++ execute tests)

[Phil Edwards <pedwards at disaster dot jaj dot com>]

[we oughta find a better list for this]


On Wed, Apr 11, 2001 at 10:19:30PM +0200, Gabriel Dos Reis wrote:
> Richard Henderson <rth@redhat.com> writes:
> | On Wed, Apr 11, 2001 at 08:04:04PM +0200, Gabriel Dos Reis wrote:
> | > This happens on a i686 running a SuSE-7.1.  Binutils is 2.10.91 (I'm
> | > not sure that is relevant though)
> | 
> | Ok, I've just bootstrapped again without a problem.
> | For reference, I've got 
> | 
> | binutils-2.10.0.18-1
> | gcc-2.96-69
> 
> I'm using gcc-3.0 from 20010408.

Just for more data points:  I've had no problems building trunk or branch
lately, on i686 running RH7, using

  gcc-2.96-54
  binutils-2.10.0.18-1
  glibc-2.2-9

as the bootstrapper.  HOWEVER, note that:

 -  more-or-less-current CVS binutils is always built along with the
    compiler, otherwise various bad things happen.  Also gdb is built
    so that I can demangle c++ coredumps.

 -  the linkonce patch disabled here

      http://gcc.gnu.org/ml/gcc-patches/2001-01/msg01679.html

    must be re-enabled; with it disabled I get segfaults during init
    and fini.  I have no idea even where to begin looking there; I am
    not a back-end expert.  I've concluded that it's local weirdness
    somehow. (how?)


> Nothing really particular.  My configure command line is:
> 
>    /codesourcery/egcs/configure --prefix=/codesourcery --enable-languages=c++ --disable-nls

/home/pme/src/3unified/configure
    --prefix=/home/pme/build/install-2001-04-11-3branch
    --with-gnu-as --with-gnu-ld --with-gcc
    --enable-languages=c++
    --enable-install-libiberty                     // always testing my patch
    --disable-nls --disable-sim --enable-gdbtk     // also building GDB
    --enable-shared --enable-debug

The --enable-debug is intended for libstdc++; I don't know what else might
be picking it up.  (The drive containing the sources is offline right now
for some TLC&PM, or I'd grep for it.)


Problems bootstrapping the branch on sparc-sun-solaris2.8, yes, but i686, no.

-- 
pedwards at disaster dot jaj dot com  |  pme at sources dot redhat dot com
devphil at several other less interesting addresses in various dot domains
The gods do not protect fools.  Fools are protected by more capable fools.