Hmm, someone (TM) should realy find a fix for our toolchain optimization
issues.

For those not following last time: when compiling a typical loop-unrolled
implementation of sha1 is compiled by our current gcc 2.95.2 toolchain
with optimization, it will take *forever* and probably fail (virtual
memory exhausted).

Compiling without optimization works. Inside libc Matt found a trick to
split the code into multiple functions that makes it compile with 
optimization. This trick doesn't seem to work when compiling mozilla.

I didn't even try it but just disabled optimization when this hit again
in the new bind version now in-tree.

Whee!


Martin