As usual, please refer to the server deployment thread for the latest updates.
There was no deployment to the Main (SLS) channel on Tuesday, February 20th, again leaving it on the same server release as weeks #6 and #7: 18#18.01.17.511913. as the channel was restarted in week #7, there was no rolling restart this week.
All three of the major RC channels should receive a new server maintenance package on Wednesday, February 21st. Release 18#18.02.12.512536 should hopefully improve (if not resolve) an odd viewer crash situation some users have experienced. At the Simulator User Group meeting on Tuesday, February 20th, Simon Linden described it thus:
The server is doing some better checking on update data it sends to the viewer. We saw a very odd situation a week or two ago where the region was sending odd data and viewers would crash immediately. It went away after we restarted the region, and we think it was some memory corruption … FWIW, the server was sending a value of zero for a prim-code … which is totally invalid … There were also some other invalid data (like a zero’ed UUID) so my theory was memory corruption.
We didn’t have any other smoking guns. That region was fine after restarting, or when we tried our own copy. It was one of those mystery bugs, which we sometimes get since SL is so big and complex. We don’t know why it got that way, or how to make it happen again. we ended up making both the region and the viewer more robust. The underlying problem is still there and, assuming it happens again, will still cause problems.
(See also: BUG-214564.)
There have been no updates to the viewer in the current official pipelines thus far, leaving them as per the end of week #7:
- Current Release version 184.108.40.2062121, dated January 26, promoted February 7 – formerly the Voice Maintenance RC.
- Release channel cohorts:
- Project viewers:
- Linux Spur viewer, version 220.127.116.119906, dated November 17, 2017 and promoted to release status 29 November – offered pending a Linux version of the Alex Ivy viewer code.
- Obsolete platform viewer version 18.104.22.1680847, May 8, 2015 – provided for users on Windows XP and OS X versions below 10.7.
Region Crossing Issues Investigation
As noted over the last few weeks, user Joe Magarac (animats) has been digging into the viewer code handling region crossings in an attempt to improve avatar handing when seated on objects and looking at the “partial unsit” issue (when the avatar becomes visual detached from a vehicle on a region crossing, but acts as if still attached (e.g. appearing seated, with any attempt to stand causing a viewer crash. Information pertaining to his effects can be found at the following location:
He now believes he has an extrapolation fix for unsits at region boundaries, which could be appearing in a future Firestorm release.
In addition, he believes he has now isolated the cause of the “partial unsit” issue as being a network bottleneck issue, and is confident he can recreate the problem simply by “overloading” his network connection by running multiple net-intensive operations in the background (resulting in packets being lost or arriving out-of-order), or by forcing packet loss.
Rather than using RLV(/a) to address this problem as a workaround, he’s now looking at using a “scripted seatbelt” – essentially a scripted attachment which can detect a partial unsit, and teleport the avatar to the last known “good” position for the vehicle, attempting to deliver the avatar 3m above the vehicle, which might make it possible for the user to then re-sit. It’s not a total solution, particularly if the vehicle has been handed-off OK and is continuing along its path, but as Simon Linden noted, at least it puts the avatar (hopefully) in the vicinity of the vehicle. And as was also acknowledged in the meeting, anything more direct is likely going to require the Lab find resources to bang on the region crossing code in both simulators and in the viewer.