testsuite oddity

[Marc Espie]

There is an annoying `feature' of the testsuite. If by any chance you get
a regression in the libio/libstdc++ tests, the corresponding call to runtest
actually returns an error...  this does stop a make check dead in its tracks.

It might be a good idea to change that: a make check is the kind of things
you start before going to bed on some machines, and you expect to read the
report in the morning.
Maybe the regressions are trivial and should be  fixed before going on,
but then maybe they're not, and the person running the test is just trying
to send regular reports to this list... this might discourage him.