[PATCH, committed] PR/41812: revert wrong multiple definitions optimizations

[Eric Botcazou]

> Yes, however, it was incorrect before.  I have an idea about how to restore
> the optimization, but I'm not sure it works and I wanted to fix the
> wrong-code first.

Then you might want to also revert the MD patch in the meantime, this cuts the 
peak memory usage in half (~550 MB vs ~1.1GB) for the same file; not being 
able to bootstrap the compiler in less than 1 GB is a little annoying.

-- 
Eric Botcazou