>To the best of my knowledge glibc has never had less then three test >failures on ARM. I've become used to ignoring them, which probably is not >the best thing. One or two of them are related thought to be compiler >bugs, which may go away with GCC 2.97.x, the head of the tree.
I thought we got down to zero failures at one point with the 2.1.x glibc, but maybe I imagined that. The wcscoll failure is indeed due to a compiler bug. I've been playing with a backport of Nick's patch to 2.95.2 which seems to fix it. I can generate a diff if anybody else wants to have a go. >Although it does look as if the 'problem' has been found and a possible >solution proposed, so hopefully it will be back again soon. Actually all >that is broken is the C++ support, the C compiler should be 'fine'. It's strange. I haven't been able to bootstrap at all for a month or so - I always get errors in `make compare'. Only one other person seems to be seeing this. I wonder if it's somehow related to the Debian environment. p.
