Christoph Hellwig wrote:
> On Sat, Feb 14, 2009 at 01:58:24PM -0600, Eric Sandeen wrote:
>> Good plan, couple comments:
>>
>> should quick only be a subset auto tests or are they orthogonal?
>
> Just a subset of auto, yes.
>
>> A few tests you've flagged as quick don't met the 20s criteria on my
>> box, just FWIW; may be filesystem-size specific:
>>
>> 013 57
>
> We need this one for sure.
agreed
>> 028 23
>> 031 21
>> 049 36
>> 075 36
>> 076 56
>> 078 20
>> 091 26
>> 112 38
>> 113 55
>> 178 363
>>
>> Most are probably close enough, but 178 looks nasty - seems to be
>> xfs_repair that's taking so long. These are running on 100G partitions,
>> though :)
>
> 178 only takes 16 seconds for me, running on 10G partitions.
interesting that a 10x size growth makes a 22x time growth.
> I'm fine removing tests, and 178 would be the first candidate, maybe 076
> and 113, too.
Fine by me to leave them in, too, not a big deal - though maybe not 178.
Just wanted to point out that some of this is probably
fs-size-dependent. It'd be neat to have this somehow run based on
what's in the timestamp file, though that requires an initial seed.
I'd just drop 178 and call it good for now.
-Eric
|