A moment of silence, please.
As of today, there has been no auto-magic regression testing for a month.
http://gcc.gnu.org/ml/gcc-regression/2002-09/
http://gcc.gnu.org/ml/gcc-regression/2002-10/
I think the quality difference in net gcc is obvious. For instance, all
targets were completely broken today for C++. This hasn't happened in a
long time, and reminds me of the scary bad days, when things were broken
more than they were fixed, and picking a time to do cvs updates was a
black art. Mistakes happen, that's not the point: what was nice about
the autobuilder was getting email hours after the offending check-in,
and the automatic checks for multiple targets, including embedded ports.
Other people's opinion of our former colleague may differ. Yes, the
communication was one-way. Yes, the messages were sometimes long, the
ChangeLogs were on occasion allowed to grow without bounds and become
savagely truncated, the subject headings were unvarying, and to some,
cold and impersonal.
I, however, remain impressed. You are sorely missed, autobuild. May you
rest in peace.
-benjamin