Re: [PATCH RFC] fstests: allow running custom hooks

[Qu Wenruo <quwenruo.btrfs@gmx.com>]

On 2021/7/19 下午10:02, Theodore Y. Ts'o wrote:
> On Mon, Jul 19, 2021 at 03:13:37PM +0800, Qu Wenruo wrote:
>> This patch will allow fstests to run custom hooks before and after each
>> test case.
>
> Nice!   This is better than what I had been doing which was to set:
>
> export LOGGER_PROG=/usr/local/lib/gce-logger
>
> ... and then parse the passed message to be logged for "run xfstests
> $seqnum", and which only worked to hook the start of each test.
>
>> diff --git a/README.hooks b/README.hooks
>> new file mode 100644
>> index 00000000..be92a7d7
>> --- /dev/null
>> +++ b/README.hooks
>> @@ -0,0 +1,72 @@
>> +To run extra commands before and after each test case, there is the
>> +'hooks/start.hook' and 'hooks/end.hook' files for such usage.
>> +
>> +Some notes for those two hooks:
>> +
>> +- Both hook files needs to be executable
>> +  Or they will just be ignored
>
> Minor nit: I'd reword this as:
>
> - The hook script must be executable or it
>    will be ignored.
>
>> diff --git a/check b/check
>> index bb7e030c..f24906f5 100755
>> --- a/check
>> +++ b/check
>> @@ -846,6 +846,10 @@ function run_section()
>>   		# to be reported for each test
>>   		(echo 1 > $DEBUGFS_MNT/clear_warn_once) > /dev/null 2>&1
>>
>> +		# Remove previous $seqres.full before start hook
>> +		rm -f $seqres.full
>> +
>> +		_run_start_hook
>
> I wonder if it would be useful to have the start hook have a way to
> signal that a particular test should be skipped.  This might allow for
> various programatic tests that could be inserted by the test runner
> framework.

Currently it's impossible, as the design is to prevent any hook to
interrupt the test.

But if we allow test case to return its result, then it should be not
hard to make us to skip test cases using start hook.

I can enhance the next version to do that, but that also means any error
inside the hook will bring down the whole test run.

Not sure the trade-off is worthy then.

Thanks,
Qu
>
> (E.g., this is the 5.4 kernel, we know this test is guaranteed to
> fail, so tell check to skip the test)
>
> 	      	      	       	    	- Ted
>