Lately I've been seeing many unexpected test case failures when
running gcc-4.4.1's test suite on one specific machine. The
gcc-4.4.1-RC's test suite worked fine on that machine a few weeks
ago, and gcc-4.4.1's works on two other machines, so this was a
bit of a puzzle.
Until I realised that the problematic machine is running 2.6.31-rc4,
and in the working cases all machines are or were running 2.6.30.
Rebooting into 2.6.30 on the problematic machine makes gcc's test
suite work on it again.
Preliminarily I have to suspect that the {p,t}ty changes break `expect'.