> I'm appalled that we seem to have hacked ld to Hell and back to work
> around a bug in gas.  *sigh*

This is the rule rather than the exception.

For commercial toolchains the problem is _much_ worse because people are
usually under pressure to get things working, but they have little or no
experience with the code base they are stuck maintaining. So they fix one
case and break others, and never find out about it until much later. But
by then, other people have already hacked their components to work around
the bugs introduced by the first fix, and the bogosity propagates around.

Between commercial toolchain makers is even worse. Since they can't
track each other's internal hacks in real time, they are stuck being
compatible with the n-1 or n-2 release of the other guy's toolchain.
Anyone doing bleeding edge work and trying to share things between
toolchains will get the death of a thousand cuts from all the little
discrepancies.

There is hope for ELF because the ABI committees for PPC/ColdFire/mcore
appear to have hired a contractor to develop an OMF test suite; I forget
who it is but the people at www.esofta.com are involved somehow. It's been
a while since I was at one of the ABI meetings.

Todd Whitesel
toddpw @ best.com