[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Easy but valuable CI task



Hi,

I think an easy but valuable way to contribute to the ROCm CI [1] would
be check the results and act upon them, where necessary.


Failing tasks
=============

If you look at Status -> Status failing tasks, you can see that we have
a large number of them.

Here, for example, for the gfx1100 architecture [2].

These need to be triaged, with bugs filed where necessary.

Sometimes a bug is just some transient issue, for example the testing
rename from trixie -> forky caused a few hiccups. These tests should be
rescheduled. This can be done via the 'recycling' symbol on the right.

Sometimes a bug is an actual bug. This must be filed as RC to prevent
the package from migrating to testing (just like a failure on the
official CI would). At least for the GPU architectures officially
supported by upstream.

The one thing that's a bit difficult to track is experimental.
Unfortunately, those tests don't show up in the [2] list.

Temporary failures
==================

If you look at Status -> Status alerts, you can see that a number of
tests failed to execute with 'tmpfail'. This is a testbed error.

These can often be solved by simple re-scheduling the task, again via
the 'recycling' symbol.

If the error persists, then the worker operator has to be notified
to resolve the issue.


A CI whose results are not acted upon, is worthless. This is an easy
entry point for contributors, and a great way to improve the quality
of our packages.

Best,
Christian

[1]: https://ci.rocm.debian.net
[2]: https://ci.rocm.debian.net/status/failing/?arch%5B%5D=amd64%2Bgfx1100
[3]: https://ci.rocm.debian.net/status/alerts/


Reply to: