There was no CCIIUG meeting this week. I think that happened because of login problems on Tuesday.
Test Clothes
Oz Linden has received a couple of clothing items for testing the Mesh Deformer. We need more items submitted. To submit items check out the original request for test clothes is: Mesh Garments needed for testing deformer. If you know someone that is making mesh clothes, ask them to submit something.
If the Contribution Agreement is holding you back, you can put items for testing into the Market Place (L$0). If you choose to go that way, include the viewer version information of the viewer used to make the upload in the item description. (Help -> About…) Then send a link to Oz Linden.
These items need to be free so that other residents can get them and use them for their testing. The CA is required so that the Lab can hand out the test items. If you are handing them out, the CA is NOT needed.
Viewer 3.4.x
The current Linden release viewer 3.4.0 version does not seem to have the memory leak the Lab has been chasing down. It is unclear whether the 3.4.1 beta viewer has the leak or not. It’s crash rate is high. Oz Linden says they are having problems reproducing the crashes, meaning there is no a single thing that crashes the viewer. Without a step-by-step to a crash it makes it hard to track down the problem.
Those playing with the viewer have noticed that it is not releasing memory when crossing to a new region. So, the more region crossings one makes the more likely they will crash. But, crowded regions seem to crash the viewer too.
Programs use a process to reserve and release computer memory. The processes are built into a set of programming functions. The functions the Lab has been using seem to have some problems releasing memory, which we call a memory leak. The result is the viewer keeps reserving more and more memory for its tasks while thinking it has release previously used memory. Since it is not being released and your computer has a limited amount of memory the viewer eventually hits a wall and crashes.
If you are wondering why memory is being reserved and released, think of coming into a region and seeing a house. The viewer realizes it needs to render the house, it requests the house data (prims and their locations) and textures. It needs to hold the information in memory, so it reserves memory and puts the info there. When you leave the region that information no longer needed, so it can release the memory.
All this reserving and releasing is done so the viewer does not have to keep track of which memory can be reused. When it is done, it turns it loose. That ‘tracking memory’ work has to be done by the Windows, Mac, or Linux for the computer in general. So, no point in duplicating that work in the viewer.
The Lab is changing to a new set of processes for reserving and releasing memory. So, far that does not appear to have fixed the problem. Of course if the programmer doesn’t set up the processes to release memory, it doesn’t matter which memory handler is used. So, somewhere a release step is likely being missed.
The high crash rates are slowing down viewer development. Actually the pipeline of fixes and features is backing up. So, we should see a number of viewer releases coming out in quick succession once the problem is resolved.
QA Fails
Oskar Linden says that the QA team has been failing several updates/fixes before they get to RC. So, if you hear someone saying the Lab doesn’t test, tell them to remember the summer of 2012.
The result of the QA testing is why there were no RC’s in the RC channels last week (38) and this week (39). So, this Tuesday will be the second Tuesday in a row without a roll out to the main channel.
There are however 3 candidates that are expected roll out to the RC channels Wednesday.
One RC is a fix to the existing system that resolves a problem with one of last week’s RC’s.
The other two are the RC’s that the Lab has been trying to get out for the last two weeks. One is a set of bug fixes:
- SVC-8119[c]: New mass-related llApplyRotationalImpulse error
- SVC-8146[c]: llRezAtRoot() does not set correct parameters (for sale) on rezzed object in Second Life RC BlueSteel 12.08.03.263047
- SVC-8144[c]: Estate Tools, Debug, Disable collisions stopped working.
The other is some bug fixes and the Group Edit fix.
It looks like these RC’s will make it to the RC Channels in week 40.
Pathfinding
This week (39) saw the last Pathfinding User Group meeting. The UG will be rolled into the Sever-Scripting group that meets Tuesdays and Fridays.
Locra Linden was late to the meeting because a Treet.tv recording session ran over time. Treet.tv is doing a show on Pathfinding that includes an interview with Lorca. It was recorded Thursday. The show will run Monday 8th, October at 2 PM PT (SLT). You can watch the show as it is presented in theater at Treet.tv’s studio.
You will also be able to see the show streamed from the Treet.tv web site. The video will be archived there, in case you miss the first showing.
Best Practices
The Pathfinding Best Building Practices is in the process of being updated. Lots of wiki stuff is supposed to be updated.
Future
Lorca said, “…the PF team would love to work on a robust animation system that could work seamlessly with Pathfinding, after we’ve had a few months at least to catch our breath.” I wouldn’t take that as a promise. But, it does show where their thinking is.
Summary
This week has been a slow week for news. With no roll outs to the grid and the viewer updates jammed up, not much is moving forward, at least from the users’ side. I am expecting a flurry of updates once the blocks are cleared.