https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82686
--- Comment #10 from Dennis Clarke <dclarke at blastwave dot org> ---
I had not yet tested a build of gcc previous to gcc 7.x regarding
this issue. Seemed counter intuitive to revert backwards but now
I have to agree that "something changed" again in the world of
doing a bootstrap of gcc. This seems to be happening more and more
and getting gcc to reasonably build is becoming more and more of
a black art as opposed to a science.
I will, for the sake of experimentation, go try a bootstrap of
the gcc 6.4.0 and then if that works out move on to 7.3.0.
I know that our friend Jonathan Wakely in comment 5 of this bug
report. Also there is question if this is a bug at all. I tend
to look at the existence of a bootstrap problem based on testing
across multiple platforms and I can not run into this same problem
on a sparc based Solaris 10 server wherein I do get multi-arch or
at least multi-lib successful build results. There are no test
results posted on the GCC site anymore :
https://gcc.gnu.org/gcc-7/buildstat.html
That seems to have died at least a year ago or more so it is
very difficult to search the vast noise of the maillist using
some format https://gcc.gnu.org/ml/gcc-testresults/YYYY-MM/ as
a target string and then the platform and version. Simply said
it is becoming more and more a black art.
Jonathan says " This library needs to be available for each
multilib variant" from the docs however this is not true at
all on some platforms. Certainly sparc has no issues.
I will try gcc 6.4.0 for the sake of a test and then move onwards.
Dennis Clarke
