Current tinderbox regression (runtime, x11perf)

[Dave Airlie]

On Tue, May 18, 2010 at 2:35 PM, Alex Deucher <alexdeucher at gmail.com> wrote:
> On Mon, May 17, 2010 at 11:01 PM, Chris Ball <cjb at laptop.org> wrote:
>> Hi,
>>
>>   > Have you updated the xserver and ddx as well?  Do you know what
>>   > versions you were running when it last worked?
>>
>> Yes -- git HEAD (of all xorg modules) as of "Wed May 5 14:25:34 2010
>> EDT" worked, and git HEAD as of three hours or so later crashed, and
>> git HEAD as of today crashes too.
>>
>> This doesn't seem to match up precisely with any DDX or server
>> commits, which is curious.  Do you have any r100 (or similarly old,
>> perhaps) hardware that you could try running x11perf on with git HEAD,
>> to see whether you share the hang?
>>
>> The x11perf invocation I use is:
>>
>> x11perf -repeat 3 -time 3 -{putimage,getimage,shmput,copy{win,pix}{win,pix}}{10,100,500}  -noop -{aa,rgb}{10,24,f}text -dot -rect{1,10,100,500}
>>
>> The hang is 100% reproducible.  I haven't tried reverting to check
>> that the hang disappears (to rule out e.g. a new hardware fault),
>> but that's probably a good sanity-check for me to do soon.
>
> There haven't been any changes to the ddx in that timeframe that would
> affect rn50.  I'd say it's an xserver bug perhaps?

It could be, not sure of anything that has changed in rn50 recently,
it mostly should be doing sw rendering at the moment.

Dave.