This is related to #48554, it would appear some power9 builders are configured differently.
Sometimes the test to open /dev/rtc0 is either skipped or passes, and sometime it fails. It is not clear if this test was always skipped previously (e.g the builder had no RTC device, or was prohibited from opening it, or has an RTC which supports alarms).
Are all the power9 builders provisioned and configured identically? Likewise, is it possible to tag golang.org/x/sys CI logs with more information about which builder they have run on (I am not sure if I am overlooking this one)?
The text was updated successfully, but these errors were encountered:
Looking into this further, it seems like we have one ppc64le power9 builder which isn't configured like the others. From farmers.golang.org I see a builder named go-ppc64le-p9 which doesn't set the HOSTNAME envvar. This matches the behavior of the failures on golang.org/x/sys.
@bcmills We have found that the failures from #48554 were due to this rogue p9 builder that is not set up like the others so fails in a different way. This builder is not documented and not set up like the others. This builder should be stopped/disabled so people don't waste time tracking down bugs on this builder.
We had opened a ticket with OSU and Paul asked them to delete it. It is gone now and it looks like all is working as expected.
The builders now match what is documented in x/build/env/linux-ppc64le/osuosl/NOTES.