I propose to enforce faillog to be used by *all* local tests.
Background:
We introduced faillog package to save some logs on test failures in issue 856540 .
My initial intention was that faillog is used in all local tests. When diagnosing failures, it's often unclear which logs are useful for debugging --- for example, some failures are often very obvious by a single screenshot because error messages are shown on display.
Currently faillog saves a screenshot and process list, but I expect more logs to be added later.
The problem is that current faillog API is opt-in and does not enforce to be used by all local tests. People may forget to enable it, or choose not to use it because it sounds not very useful for their particular tests --- but no one knows which logs are useful for debugging before failures happen, actually.
Challenges:
I want faillog to be invoked even if each test function does not mention it at all. However this needs changes to Tast's test runner, so we need design discussion.
Comment 1 by derat@chromium.org
, Sep 12