El 23/12/24 a las 18:18, László Böszörményi (GCS) escribió:
I do not see in your statistics how many builds were tried, ie the failing percent and not its plain number.
You are right. My mistake. The 128 build failures I put in my people.debian.org page are the end result of building the package 100 times on machines with 1 CPUs and 100 times on machines with 2 CPUs. Therefore the failure rates are 100% and 28% respectively. The funny thing here is that the way it fails 100% of the time on a system with a single CPU is very similar to the way it failed on s390x. So what it seems to me (and of course I may be wrong) is that this is not really related to number of CPUs or architecture, but maybe overall speed. The build history here: https://buildd.debian.org/status/logs.php?pkg=nng&arch=s390x may be interpreted in several ways. We could think that it used to fail and it does not fail anymore, but I think a more accurate theory would be to consider than on s390x it fails 80% of the time and it probably still fails 80% of the time. Fortunately this is high enough that we don't need to debate about how much support we want for systems outside buildd.debian.org.
OK, I will disable the mentioned tests in the next few days.
Thanks a lot.
