Aurelien Jarno wrote...
> The GNU libc has never been designed to work in a multiarch situation,
> hence this kind of bugs (and I am sure there are been plenty more). We
> can't therefore consider that kind of bug as important or critical.
Thanks for your time to write down that explanation. I understand now,
and although it's not so good news, I have to agree with your point.
> In short we will work on this bug if we have been able to fix all the
> other important bugs. That said, if someone interested by multiarch can
> provide us a patch, we will included happily (of course provided it is
> not too hackish).
Sounds challenging but unfortunately I already got way to many other
things on my list. In case I cannot resist the temptation to hack on
this: Can you think of a reason why different architectures would ever
want to share ld.so.cache as they do right now? Since I cannot, the
solution seems to qualify that file with <triple>, teach the so loader
to pick the appropriate one, and hack ldconfig to do the right thing.
Could it be that (relatively) simple?
Christoph
Attachment:
signature.asc
Description: Digital signature