Hi Gabriel, I think this is related to some interaction between GWC  
and the MockTest stuff that I don't understand well enough.

As stated below, the exception happens because TileLayerDispatcher  
doesnt have a table of available layers yet. What normally happens is  
that Spring creates the beans that represent configuration sources and  
hands them through the constructor to TileLayerDispatcher. The sources  
are not queried immediately, in case they live in the same container.  
Instead, a thread is launched that immediately acquires a lock and  
then waits for a specified number of seconds, all requests are blocked  
until this thread completes.

Normally this works fine, since the servlet shouldnt be accessible  
until Spring is done loading, but I guess the MockTest circumvents  
this and does the getLayer before before the thread acquires the lock.

I think the clean solution is to use an explicit lock object and  
acquire that during the construction of TileLayerDispatcher, then have  
the thread release it. But I don't have capacity to do a 1.2.2a  
release early this week, so I am thinking we should introduce a 10  
second delay (apologies) in the GS test and hope that solves the issue  
for now, and then we remove that delay as soon as GS gets pegged to  
the next GWC release.

But that assumes that the mock stuff has no clever tricks to deal with  
these problems, any thoughts?
-Arne


Quoting Andrea Aime <[email protected]>:

> Arne Kepp ha scritto:
>> Uh, which one of the very many builds ? :)
>>
>> Both 2.0.x and trunk have been pegged to 1.2.2 and not changed in 2  
>> months (except for the KML superoverlays patch, which I dont think  
>> has anything to do with it). So it doesn't ring any bells unless  
>> those intermittent failues have been going on for that long.
>
> They have been on and off for quite a bit of time but I'm not sure
> if it's happening only on trunk or also on 2.0.x (for sure this
> build failed on trunk).
>
>> But there is a null in the resource path that I don't think should be there.
>
> Scratch scratch... maybe it's due to the mock test setup?
> How is that path determined?
>
> Cheers
> Andrea
>
>
> -- 
> Andrea Aime
> OpenGeo - http://opengeo.org
> Expert service straight from the developers.
>





------------------------------------------------------------------------------

_______________________________________________
Geoserver-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/geoserver-devel

Reply via email to