On Thu, Apr 22, 2010 at 7:22 PM, Martin Bligh <[email protected]> wrote:
> On Thu, Apr 22, 2010 at 11:54 AM, Akshay Lal <[email protected]> wrote:
>> Well previously, when the fsck ran on an unmounted file system and the file
>> system was dirty/corrupted in anyway, the fsck would throw an error (but fix
>> the file system '-fy'). This error would get logged in the
>> console/tests-specific log as an 'ERROR', but the overall test would pass.
>> Thats not really very non-intuitive and not correct (either) since you
>> ideally want the file system in a clean state after all the tests run.
>> Explicitly stating that a failure in an fsck run results in a test failure
>> seems like the right approach - I think.
>
> That seems odd. I'd think the test level error handler would convert
> it. Greg - do you recall from refactoring that code?
>
> I'm suspicious of throwing a TestError from something that can
> be run from outside of a test.
>

I suggested that Akshay put this here.  But if thats wrong, we need to
look closer to figure out where it goes.

What calls this code?  The only caller I see is
partition.cleanup_after_test() which is called by
partition.run_test_on_partition after job.run_test() is finished but
still within the context of job.run_group().

Is it crazy of me to think that filesystem tests should be calling
this directly within the context of a test if they want to depend on
the error being flagged as a test problem.  Which either means that
partition.run_test_on_partition probably needs to be reworked to make
sure the fsck happens within run_test context or that the tests using
it should explicitly call partition.fsck themselves when they want the
fsck result to change the test outcome.

-gps
_______________________________________________
Autotest mailing list
[email protected]
http://test.kernel.org/cgi-bin/mailman/listinfo/autotest

Reply via email to