so it sure looks like a gcc upgrade caused the failure. But it's not
clear wheter it's a compiler bug, or some undefined behaviour that
triggers the bug.
Fabien, any chance to either bisect or get a bit more information on
the backtrace?
There is a promising "keep_error_builds" option in buildfarm settings,
but it does not seem to be used anywhere in the scripts. Well, I can
probably relaunch by hand.
However, given the experimental nature of the setup, I think that the
most probable cause is a newly introduced gcc bug, so I'd suggest to
wait to check whether the issue persist before spending time on that,
and if it persists to investigate further to either report a bug to gcc
or pg, depending.
Also, I'll recompile gcc before the next weekly builds.
I did some manual testing.
All versions are tested failed miserably (I tested master, 12, 11, 10,
9.6…). High probability that it is a newly introduced gcc bug, however pg
is not a nice self contain tested case to submit to gcc for debugging:-(
I suggest to ignore for the time being, and if the problem persist I'll
try to investigate to detect which gcc commit caused the regression.
--
Fabien.
