Re: 3.0, sun, R_SPARC_32 failure

[Alexandre Oliva <aoliva at redhat dot com>]

On Jun  7, 2001, Mark Mitchell <mark@codesourcery.com> wrote:

>>>>>> "Nathan" == Nathan Sidwell <nathan@codesourcery.com> writes:
Nathan> ok for mainline & branch?

> Yes, thank you.

Actually, this is far from being ok.  For a start, it relies on
dirname, that is not portable.  Second, it assumes gcc_ac_as is set;
it needs not be.  Third, it assumes readelf and/or elfdump are in the
same directory as gas, which is not the case in a unified build tree.
Fourth, it assumes readelf is named readelf, and not
target_alias-readelf, as it's going to be named in an installed cross
tool-chain.  Fifth, it fails to take into account that, in a
unified-tree build, the assembler won't have been built at GCC
configure time, so you have to peek at its version number.

Some of these problems have broken my native unified-tree builds to
fail on Solaris 2.5/sparc and Red Hat Linux 6.2/sparc.

Since this patch doesn't affect the functionality of GCC at all, and
it actually breaks bootstrapping, I suggest that we invoke the
broken-bootstrap rule and revert the patch.  Alternatively, we could
downgrade the ERROR it prints to a WARN followed by sleep 30, or make
it overridable by a command-line switch (actually, it's enough to set
gcc_cv_as_ua32_reloc=yes in configure's environment, but do we want to
document it?)

-- 
Alexandre Oliva   Enjoy Guarana', see http://www.ic.unicamp.br/~oliva/
Red Hat GCC Developer                  aoliva@{cygnus.com, redhat.com}
CS PhD student at IC-Unicamp        oliva@{lsd.ic.unicamp.br, gnu.org}
Free Software Evangelist    *Please* write to mailing lists, not to me