[Opensim-dev] Fwd: Issue 1096 in singularity-viewer: [AVFACTORY]: Missing baked texture spamming console

[Justin Clark-Casey]

This is an issue if all caching is disabled.  However, it should have been fine with just disk caching enabled.  This is 
the default that OpenSimulator ships with - one shouldn't need to do any copying of FlotsamCache.ini.example, though it 
won't hurt.

On 14/10/13 16:12, Fleep Tuque wrote:
> I'm not sure if there is already a mantis bug report on this and I don't have time right this second to look, but
> thought I'd forward this message along as an FYI.  I recall having some discussions about this with Justin and Neb prior
> to OSCC but I don't recall what the resolution was if any.
>
> - Chris/Fleep
>
> ---------- Forwarded message ----------
> From: ** <singularity-viewer at googlecode.com <mailto:singularity-viewer at googlecode.com>>
> Date: Mon, Oct 14, 2013 at 10:35 AM
> Subject: Re: Issue 1096 in singularity-viewer: [AVFACTORY]: Missing baked texture spamming console
> To: fleep513 at gmail.com <mailto:fleep513 at gmail.com>
>
>
>
> Comment #1 on issue 1096 by jayden.b... at gmail.com <mailto:jayden.b... at gmail.com>: [AVFACTORY]: Missing baked texture
> spamming console
> http://code.google.com/p/__singularity-viewer/issues/__detail?id=1096
> <http://code.google.com/p/singularity-viewer/issues/detail?id=1096>
>
> As a very interested observer of this issue, but not a Singu team member just a user, I would like to make a hopefully
> helpful comment and also confirm this issue.
>
> I noticed this issue about 2 weeks ago and have been testing extensively since then as I too suffered it.
>
> I have mostly solved the byproduct; the error message in the log (not the root cause) by enabling the server cache (copy
> the FlotsamCache.ini.example to FlotsamCache.ini - OS setup documents mostly forget to constantly tell people this
> rather important step >.>).
>
> As a side effect loads are also much faster.
>
> Googlefu found and confirmed that enabling *both* memory and disk caching in the FlotsamCache.ini file must be done to
> remove the error. (Devs: Remind me to add the cross reference to OpenSim's mantis later. hehe.)
>
> With 2 weeks solid testing, it seems at least on my OS regions if you don't have both settings set to on then the baked
> textures are obtained by the server on TP in, but not applied and then lost by the OpenSim server as it deals with
> everything else before finally rendering you. Before this workaround, on my regions at least, when it came time to
> render the avatar I saw the warning in the logs. The previously sent but lost (due to insufficient server caching) baked
> textures are thus re-requested and applied and you finally turn from cloud to avatar.
>
> (As a hint to the reporter, I also bumped up my cache retention quite a lot as my three regions are not so high traffic.
> I can afford the ram and disk to hold things for 3 days to a week even. YMMV depending on scenes and visitors.)
>
> All that said:
>
> There is STILL the root cause to this issue as I see it, in that the baked textures for a new person to the region - to
> my inexpert eyes looking at both server and client logs in depth - are done at the wrong interest level or lost by the
> server. As a hint to any developer who looks at this in the future; it's almost looks like the avatar needs to be baked,
> rendered and placed in world before the world is drawn around it... as an alternative, there could be a new step
> included in the initial region processes... draw a known default texture as skin/eyes until the baked texture is ready.
>
> I suspect Cool does this by displaying a blank white skin, but I've not yet bothered digging into Henri's code to see
> his differences. Avatars on Cool do rez super fast, but they come in as blank textured, 0xFFFFFF coloured and remain
> that way it seems for as long as Singu people are clouds. It's better to look at.
>
> ((Votes +1 for return of your shape but 0xFFFFFF skinned Ruths))
>
> No matter what viewer though... My own server logs after this flotsam change is made show new people to the region will
> be unbaked for a while. Far less than without a server cache though. Some people will need to force rebake. But the
> error is gone.
>
> As a respectful triage hint... I'd love to be proven wrong but this seems like it's an OpenSim root cause, but I do feel
> Singu could handle that failure better. I should discuss it one day in depth with a dev when new people into my region
> stay as a cloud.
>
>
>
> Reporter: If you have fcache turned on with both memory and hard disk and are seeing those errors then my findings are
> for a different bug. But as the error messages were the same 2 weeks ago and I've not had one since... it is worth
> writing this.
>
>
> /me leaves it for the very capable Singu team to research/triage further. ;)
>
> --
> You received this message because you starred the issue.
> You may adjust your notification preferences at:
> https://code.google.com/__hosting/settings <https://code.google.com/hosting/settings>
>
> Reply to this email to add a comment.
>
>
>
> _______________________________________________
> Opensim-dev mailing list
> Opensim-dev at lists.berlios.de
> https://lists.berlios.de/mailman/listinfo/opensim-dev
>


-- 
Justin Clark-Casey (justincc)
OSVW Consulting
http://justincc.org
http://twitter.com/justincc