[patch] to bug 82243 (not 88243)

[Paul Temple]

Leo Savernik wrote:
> Before I implement the changes, I update the regression
> testsuite, and run testregression, saving the output in a log.
>
> Then I implement the changes, and run testregression again,
> saving the output in another log.
>
> Then I'll diff -u oldlog newlog.
>
> If there are any differences, they may have been caused by
> your changes and must be investigated.

Hm, but then I won't find a regression if it occurs in a test 
that also failed without my changes - and there are MANY failed 
tests. Even after undoing all my local changes ;)

Do you only compare stdout before / after?
As there are quite some
    testregression: pixel (25, 45) is different #ffffff #000083
and similar printed to stderr

There are so many FAILs... is this just temporary? If not, what 
are the regression tests good for?