Hello Tom,
>>> Could you please just shut down the animal until that's dealt with?
>
>> The test is failing because there is a problem, and shuting down the test
>> to improve a report does not in any way help to fix it, it just helps to
>> hide it.
>
> Our buildfarm is run for the use of the Postgres project, not the LLVM
> project.
The point of these animals is to have early warning of upcoming compiler
changes. Given the release cycle of the project and the fact that a
version is expected to work for 5 years, this is a clear benefit for
postgres, IMO. When the compiler is broken, it is noisy, too bad.
In this instance the compiler is not broken, but postgres is.
If the consensus is that these animals are useless, I'll remove them, and
be sad that the community is not able to see their value.
> I'm not really happy that it contains any experimental-compiler
> animals at all, but as long as they're unobtrusive I can stand it.
> serinus and seawasp are being the opposite of unobtrusive.
I think that the problem is the report, not the failing animal.
In French we say "ce n’est pas en cassant le thermomètre qu’on fait tomber
la fièvre", which is an equivalent of "don't shoot the messenger".
> If you don't want to shut it down entirely, maybe backing it off to
> run only once a week would be an acceptable compromise. Since you
> only update its compiler version once a week, I doubt we learn much
> from runs done more often than that anyway.
Hmmm… I can slow it down. We will wait one week to learn that the problems
have been fixed, wow.
<Sigh>.
--
Fabien.