As noted in my week 14 (2) update, prims missing from linksets / builds is not a new issue, it’s been going on for some months now. In my own case, I’ve found that on teleporting to my home, an internal wall I’ve added to the build is routinely missing from my view. Right-clicking on the area of the wall has caused it to immediately render. As the wall in question tends to rez OK when I log-in to SL, I’ve tended to look on the matter as a viewer issue.
However, following the week 14 deployments, the incidences of missing linksets seems to have both risen in both numbers and size of “missing” prims. There have been numerous reports on the deployment discussion thread, including questions as to whether it is a viewer issue or server issue, and things seem to have become a lot more visible.
My personal experience has been that things are worse – as shown in the image above. Prior to the April 2nd, The wall on the far side of the house, with the two pictures and turning to run under the bedroom balcony rails, would fail to render following a teleport. Since the Main channel deployment on Tuesday April 2nd, that wall now renders flawlessly – but a huge section of the linkset for the house itself is almost constantly missing following a teleport home. Camming around my home region reveals other houses in the same condition.
Such has been the level of discussion on the thread about the matter, lead by Wolfbaginski Bearsfoot and others, that Maestro Linden stepped-in and posted some pointers to help people determine if they have a viewer or a server-side issue, stating:
Hi Wolfbaginski, I would do the following to determine whether objects are missing due to a server bug or viewer bug:
1) Note whether the missing objects are whole linksets or only certain prims in linksets. Interest list issues would generally affect whole linksets. A viewer rendering issue or maybe a message-packing/decoding issue would seem more likley if only certain prims in linksets are affected. It sounds like you’re seeing at least part of the linkset, since you’re able to select it.
2) While the prims are missing, enable Develop -> Render Metadata -> Bounding Boxes, and see what bounding boxes appear. You should see a prim-aligned bounding box for each prim, as well as a world-aligned bounding box that covers the linkset. If you see a bounding box where a missing prim is located, or if the linkset bounding box extends to include the missing prims, then it’s almost certainly the viewer’s fault.
3) While the prims are missing, enable Develop -> Show Info -> Show Updates to Objects. With this setting enabled the viewer will render a particle from each linkset/prim that it got an update for:
- Red means a full update was received from the server (which has a full description of the object’s visual parameters)
- Blue means a terse update was received (which only includes information about a few properties, such as position and velocity)
- Green means that an ‘objectdelete’ update was received (meaning that the object was either derezzed or is out of range for the viewer)
If you enable this feature, and observe that the missing prims appear without a full object update being sent, then it’s probably a viewer bug (in that the viewer knew about the missing prims the whole time, but initially failed to render them).
4) If you leave a region, then return to it, the viewer will load cacheable objects from the local cache, instead of getting the object details from the simulator. If the object was loaded from cache and the appearance has been either fixed or broken since the last time you saw it, this would indicate a viewer bug. You can verify that an object was loaded from viewer cache by enabling Develop->Rendering Metadata->Update Type; objects loaded from cache are shaded blue, with this mode enabled.
As a result of this, a healthy exchange started between users and Maestro, as Wolfbaginski noted in his feedback to my original report, with further tests being carried out. The discussion further confirmed that the problem appeared to be worse since the deployment on April 2nd, but that findings pointed to the issue being more towards the viewer end of things than the server – and that the problem was as prevalent in the recent SL viewer releases as it is in “older” TPV releases. The discussions also lead to some feedback on how to force “missing” prims to appear, including:
- Flicking alpha masks off & on
- Toggling Atmospheric Shaders in Graphics Preferences off and on.
The result of this collaboration was an update from Maestro Linden, posted to his original notes (above) which reads:
I investigated this on region LEA27 on Aditi. This region has 25k prims, which are all identical mesh boxes. They are arranged in pretty patterns, so it’s fairly obvious when some are missing. I found that some prims in linksets ‘randomly’ (different linksets are affected each time I visit the region) don’t render until selected. All of the debug settings indicate that this is a viewer-side issue. I used ‘Second Life 3.4.5 (270263) Feb 12 2013 04:36:12 (Second Life Release)’ to test. I didn’t see an existing Jira for this issue, so I filed a new one internally.
How long it might be for any fix for the issue appearing as a result of Maestro’s JIRA obviously isn’t clear at this point – and there is a lot going on with the viewer at present. However, the important aspects of this are that:
- Good collaboration between users and LL has led to an issue being positively pinned-down
- A JIRA on the matter has been raised internally within LL and will as a result be getting attention
- There are some alternative “fixes” available for the problem should you witness it (right-click on the missing prims, flick alpha masks off / on or toggle Atmospheric Shaders).
If you encounter the problem, and while it is frustrating, try one of the three solutions above.
With thanks to Wolfbaginski Bearsfoot