public inbox for linux-xfs@vger.kernel.org
 help / color / mirror / Atom feed
* xfstests: kludge patch for per-dir test number sorting by ./check
@ 2013-04-13 20:50 Michael L. Semon
  2013-04-14 23:26 ` Dave Chinner
  0 siblings, 1 reply; 5+ messages in thread
From: Michael L. Semon @ 2013-04-13 20:50 UTC (permalink / raw)
  To: xfstests

I'm trying to get the new xfstests to run the XFS tests first, then run 
the generic group, then run the shared group.  One attempt was to work 
with this variable in the ./check script:

SRC_GROUPS="generic shared"

There is a final file sort in ./check that means, no matter which order 
groups are specified in SRC_GROUPS, the groups will be executed in 
alphabetical order.  This sysadmin kludge patch fixes that behavior and 
seems to work.  By moving the sorting into the trim_test_list() 
function, test numbers are still sorted, but the groups seem to run in 
the order specified in the SRC_GROUPS variable.

Let me know if this helps or if I'm simply not using the new xfstests 
correctly.

Michael

--- xfstests/check.orig	2013-03-30 00:54:37.000000000 -0400
+++ xfstests/check	2013-04-13 13:41:42.002814733 -0400
@@ -138,7 +138,7 @@
  	    echo "^$t\$" >>$tmp.grep
  	    numsed=`expr $numsed + 1`
  	done
-	grep -v -f $tmp.grep <$tmp.list >$tmp.tmp
+	grep -v -f $tmp.grep <$tmp.list | sort -n >$tmp.tmp
  	mv $tmp.tmp $tmp.list
  }

@@ -268,7 +268,7 @@
  fi

  # sort the list of tests into numeric order
-list=`sort -n $tmp.list`
+list=`cat $tmp.list`
  rm -f $tmp.list $tmp.tmp $tmp.grep

  if $randomize

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: xfstests: kludge patch for per-dir test number sorting by ./check
  2013-04-13 20:50 xfstests: kludge patch for per-dir test number sorting by ./check Michael L. Semon
@ 2013-04-14 23:26 ` Dave Chinner
  2013-04-15  3:29   ` Michael L. Semon
  0 siblings, 1 reply; 5+ messages in thread
From: Dave Chinner @ 2013-04-14 23:26 UTC (permalink / raw)
  To: Michael L. Semon; +Cc: xfstests

On Sat, Apr 13, 2013 at 04:50:06PM -0400, Michael L. Semon wrote:
> I'm trying to get the new xfstests to run the XFS tests first, then

The question is why do you want to do this? Is there any specific
reason for running the tests in that order?

FWIW, if all you want to do is run the xfs tests, run:

$ sudo ./check xfs[0-9][0-9][0-9]

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: xfstests: kludge patch for per-dir test number sorting by ./check
  2013-04-14 23:26 ` Dave Chinner
@ 2013-04-15  3:29   ` Michael L. Semon
  2013-04-22 18:25     ` Rich Johnston
  0 siblings, 1 reply; 5+ messages in thread
From: Michael L. Semon @ 2013-04-15  3:29 UTC (permalink / raw)
  To: Dave Chinner; +Cc: xfstests

On 04/14/2013 07:26 PM, Dave Chinner wrote:
> On Sat, Apr 13, 2013 at 04:50:06PM -0400, Michael L. Semon wrote:
>> I'm trying to get the new xfstests to run the XFS tests first, then
>
> The question is why do you want to do this? Is there any specific
> reason for running the tests in that order?
>
> FWIW, if all you want to do is run the xfs tests, run:
>
> $ sudo ./check xfs[0-9][0-9][0-9]
>
> Cheers,
>
> Dave.
>

Personal preference.  I'd like to run both series of tests, sometimes 
attended, sometimes unattended.  It's fine if a generic/ test crashes 
the PC while running unattended, but I'd much rather the xfs/ tests have 
first crack at that.

My version of the syntax was `./check xfs/[0-9][0-9][0-9]`, but that 
becomes rough when skipping tests.  I'll still try your version, but 
irrationality and frustration have driven me to simply adding a "mls" 
group to all of the tests/<fsgroup>/group files, then hoping that a 
`./check -g mls` would run everything in the correct order.  That led to 
trying to figure out why the generic/ tests were running first, and so 
on and so forth...

Thanks!

Michael

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: xfstests: kludge patch for per-dir test number sorting by ./check
  2013-04-15  3:29   ` Michael L. Semon
@ 2013-04-22 18:25     ` Rich Johnston
  2013-04-23 14:08       ` Michael L. Semon
  0 siblings, 1 reply; 5+ messages in thread
From: Rich Johnston @ 2013-04-22 18:25 UTC (permalink / raw)
  To: Michael L. Semon; +Cc: xfstests

On 04/14/2013 10:29 PM, Michael L. Semon wrote:
> On 04/14/2013 07:26 PM, Dave Chinner wrote:
>> On Sat, Apr 13, 2013 at 04:50:06PM -0400, Michael L. Semon wrote:
>>> I'm trying to get the new xfstests to run the XFS tests first, then
>>
>> The question is why do you want to do this? Is there any specific
>> reason for running the tests in that order?
>>
>> FWIW, if all you want to do is run the xfs tests, run:
>>
>> $ sudo ./check xfs[0-9][0-9][0-9]
>>
>> Cheers,
>>
>> Dave.
>>
>
> Personal preference.  I'd like to run both series of tests, sometimes
> attended, sometimes unattended.  It's fine if a generic/ test crashes
> the PC while running unattended, but I'd much rather the xfs/ tests have
> first crack at that.
>
> My version of the syntax was `./check xfs/[0-9][0-9][0-9]`, but that
> becomes rough when skipping tests.  I'll still try your version, but

I think this is a reasonable change request and I can see how this would 
be very rough if you wanted to skip tests. Anyone have a reason why not 
to do this?

Regards,
--Rich

> irrationality and frustration have driven me to simply adding a "mls"
> group to all of the tests/<fsgroup>/group files, then hoping that a
> `./check -g mls` would run everything in the correct order.  That led to
> trying to figure out why the generic/ tests were running first, and so
> on and so forth...
>
> Thanks!
>
> Michael

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: xfstests: kludge patch for per-dir test number sorting by ./check
  2013-04-22 18:25     ` Rich Johnston
@ 2013-04-23 14:08       ` Michael L. Semon
  0 siblings, 0 replies; 5+ messages in thread
From: Michael L. Semon @ 2013-04-23 14:08 UTC (permalink / raw)
  To: Rich Johnston; +Cc: xfstests

On 04/22/2013 02:25 PM, Rich Johnston wrote:
> On 04/14/2013 10:29 PM, Michael L. Semon wrote:
>> On 04/14/2013 07:26 PM, Dave Chinner wrote:
>>> On Sat, Apr 13, 2013 at 04:50:06PM -0400, Michael L. Semon wrote:
>>>> I'm trying to get the new xfstests to run the XFS tests first, then
>>>
>>> The question is why do you want to do this? Is there any specific
>>> reason for running the tests in that order?
>>>
>>> FWIW, if all you want to do is run the xfs tests, run:
>>>
>>> $ sudo ./check xfs[0-9][0-9][0-9]
>>>
>>> Cheers,
>>>
>>> Dave.
>>>
>>
>> Personal preference.  I'd like to run both series of tests, sometimes
>> attended, sometimes unattended.  It's fine if a generic/ test crashes
>> the PC while running unattended, but I'd much rather the xfs/ tests have
>> first crack at that.
>>
>> My version of the syntax was `./check xfs/[0-9][0-9][0-9]`, but that
>> becomes rough when skipping tests.  I'll still try your version, but
>
> I think this is a reasonable change request and I can see how this would
> be very rough if you wanted to skip tests. Anyone have a reason why not
> to do this?
>
> Regards,
> --Rich

It's a good change request, but on further testing, it seems like my 
kludge patch breaks basic sorting, i.e. `./check generic/003 generic/001 
generic/002` will run the tests in exactly that order.  That's not 
necessarily bad--principle of least surprise--but it means that to get 
the old xfstests sorting behavior, it would have to be sorted using 
another method.

Thanks for considering the need for more control over the order in which 
groups are run.  It's most certainly appreciated.

Thinking out loud:  Is there a disadvantage to setting $SRC_DIR to "." 
instead of "tests"?  It looks like a nice way to get `./check 
tests/generic/001` to run tests/generic/001 instead of issuing "unknown 
test, ignored", but I've tested it only for generic/001 here at work.

Michael

>> irrationality and frustration have driven me to simply adding a "mls"
>> group to all of the tests/<fsgroup>/group files, then hoping that a
>> `./check -g mls` would run everything in the correct order.  That led to
>> trying to figure out why the generic/ tests were running first, and so
>> on and so forth...
>>
>> Thanks!
>>
>> Michael

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

^ permalink raw reply	[flat|nested] 5+ messages in thread

end of thread, other threads:[~2013-04-23 14:08 UTC | newest]

Thread overview: 5+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2013-04-13 20:50 xfstests: kludge patch for per-dir test number sorting by ./check Michael L. Semon
2013-04-14 23:26 ` Dave Chinner
2013-04-15  3:29   ` Michael L. Semon
2013-04-22 18:25     ` Rich Johnston
2013-04-23 14:08       ` Michael L. Semon

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox