[Xcb] Problem selecting event input from different threads.

[Erik De Rijcke]

I've tested with the xlib-xcb-thread-fixes. Thread A never receives any
configure requests or map request events, while another thread B is waiting
eternally for a reply. This is basically the same behavior as originally
described with the difference that thread B now locks. Thread B is locked
before an event is send to thread A and keeps being locked while thread A
should have received an event.

What's perhaps also interesting to know is that a substructure redirect mask
is set on the root window. This should keep client windows from being made
visible and events should be send, yet they are instantly mapped like no
mask is set on the root window, hence why thread A never receives any
events. Other trivial events, like KeymapNotify, still arrive.

On Mon, Apr 19, 2010 at 12:15 PM, Vincent Torri <vtorri at univ-evry.fr> wrote:

>
>
> On Mon, 19 Apr 2010, Jamey Sharp wrote:
>
>  On Mon, Apr 19, 2010 at 2:56 AM, Erik De Rijcke <derijcke.erik at gmail.com>
>> wrote:
>>
>>> I'll try to make a simple unit test (Java with C) that replicates
>>> this behavior if it would help squashing this bug?
>>>
>>
>> It certainly improves our chances!
>>
>
> unit testing (there are some but too few) + benchmarks should have been
> done since a long time. Profiling too. Maybe it's time to do them...
>
> Vincent
>
>
>>  I've tested with the latest xlib and xcb from git master.
>>> These are the results I'm getting.
>>> Xcb: same results as described earlier
>>>
>>
>> Darn, I thought the patch on libxcb master would help there.
>>
>> For your Xlib testing, would you try the xlib-xcb-thread-fixes branch
>> please?
>>
>> Thanks,
>> Jamey
>> _______________________________________________
>> Xcb mailing list
>> Xcb at lists.freedesktop.org
>> http://lists.freedesktop.org/mailman/listinfo/xcb
>>
>>
>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/xcb/attachments/20100419/d7a4df12/attachment-0001.htm>