The following notes are primarily taken from the TPV Developer (TPVD) meeting held on Friday, February 12th, 2016. A video of the meeting is embedded at the end of this report, and time stamps in the text relate to that recording. My thanks as always to North for supplying the video.
Server Deployments – Recap
There were no server deployments for week #6, as the majority of staff at the Lab directly involved in developing and enhancing the platform were meeting to discuss plans for further enhancements and improvement to Second Life.
This more than likely means the next deployment will take place Wednesday, February 17th, 2016, to at least one of the RC channels.
[00:19] The current version of the HTTP / Vivox viewer (184.108.40.2060660) has HTTP pipelining disabled as a result of bug within it which was causing the viewer to crash. The Lab is now looking to fix the underlying cause of the issue.
[00:55] The Quick Graphics viewer is undergoing a further round of changes to refine how it performs the avatar complexity calculations.
[02:40] It is thought that all of the remaining issues in this regard (such as blocked avatars not rendering correctly, elevated costs involved when using alpha masking instead of alpha blending and materials are not being accounted for correctly.
[03:22} Further changes to the viewer mean that from the next release it will no longer use a LOD of 3 on which to base avatar complexity calculations, but will be more responsive to the different LODs set within a mesh, depending on which you are seeing
So, for example, this means if you are close enough for the high poly version of the mesh to be rendered, the Avatar complexity calculation will be based on that; if you are further away, and a lower polycount version is rendered, then the LOD for that will be used in calculating the Avatar complexity.
Once all of these updates have been incorporated into the viewer, which should be in the next week or two, an updated RC will be issued, paving the way for this viewer to be promoted as the de facto release viewer sometime in the not too distant future after that.
An interesting point with this Quick Graphics viewer is that those using the RC version are tending to run longer log-in sessions which are “significantly longer” than with those users running the current release viewer.
The Oculus Rift project viewer is still anticipated as appearing “soon”, and will lift this viewer to being compatible with the latest SDK from Oculus VR, but will be for Windows only, given the Oculus SDK currently doesn’t support Mac.
64-Bit Viewer Versions
[05:00] The viewer team continue to pursue 64-bit versions of the official viewer, but it will still be a while before any initial versions are visible and available for people to try. It is likely that any project viewer will not be appearing for a “few weeks”. Once project viewers are released into the wild, it a likely the Lab will make available 64-bit versions of the pre-built Havoc extensions libraries to TPVs with the licence as well. The release of the 64-bit viewers will also see the Lab cease support for 32-bit versions of OS X, something which should hardly affect any Mac users.
[01:20] With the Bento Bones survey closed and feedback also gathered through meetings, it is hoped that the skeleton changes can be finalised relatively soon, allowing the Lab to focus on issues of bug fixing within the viewer and readying it for a release (project or C) on the main (Agni) grid. This will also see the server-side support for Bento enabled on Agni, having been deployed last year as a part of the routine server releases (week #48 2015 for the Main (SLS) channel and originally to the RC channels in week #46 of 2015).
[06:46] The Lab, with content creators, is continuing to experiment with reconfiguring the bones, and also to look into the issues of joints getting out of position. As noted in my Bento Project reports, there are a number of different circumstances where this can occur, some of which have yet to be fully diagnosed, and others of which could be caused by race conditions between the viewer and the server when it comes to message handling, or which may even be due to message packets being lost for those on poorer network connections.
A further issue is to do with the default human T-pose attempting to load between animations, which can cause quadruped avatars appear to try to cross their forelegs, depending on the bones used in rigging them. For a details discussion on all these issues, see my Project Bento update #4, with audio.
CEF, and Firestorm and Anti-Virus software
[14:08] One of the significant factors in delaying the next Firestorm release is the merge with the Chrome embedded Framework (CEF) code which is now used for media handing, etc., within the viewer. A set of fixes for CEF are already available in the current Maintenance RC viewer (version 220.127.116.110545 at the time of writing), with further updates expected in the next Maintenance RC thereafter – although none of these are considered major issues.
[15:43] There have been reports that one or two anti-virus packages (one of them being Webroot) don’t play nice with CEF versions of the viewer. There’s not a lot the Lab can do about this, and users experiencing problems may want to look into the AV software and see if the viewer requires whitelisting, etc.
Find out more on this through my TLS 1.2 implementation notes.
Avatar Issues and Interest List (?)
[17:21] There have been increasing reports over the last couple of months of avatar location discrepancies occurring over the last few months. Examples of this are avatars standing in front of you failing to render (including tags), although you can see them on radar, read their local chat and hear them on voice. Sometimes they might be reported at 0,0,0 within the region, or seem to be a few hundred metres up in the air, etc.
There is a belief that these problems may be Interest List related, although the Lab hasn’t intentionally made any Interest list changes in the period these issues have been occurring.
The recommendation is that if this happens to you, logout out of Second Life, grab your log files, and file a JIRA on the issue. When doing so, include as much information as possible – date, time, region, a description of what happened, any snapshots / screen captures you made prior to logging-out, etc. Please try to ensure you do this ASAP after the event, as the server logs, which would be used to see if there is anything untoward on the server which might account for the issue, are generally only retained for a few days.
Better yet would be keeping an eye on when this occurs (assuming it happens to you more than once or twice), and seeing if there is any correlation between where / how it happens, etc (e.g. does it seem to happen when you are specifically teleporting from X to Y with someone, or when you enter a specific region, and so on), which may give a clue as to how the problem can be reliably reproduced and thus investigated.
REMINDERS: TLS 1.2 Implementation and Open-source Development Meeting
A further reminder was given that the Lab will be switching over to using TLS 1.2 in the near future (possibly spring 2016). The code is already in the viewer, and is being adopted by TPVs. When the switch-over happens, it means anyone using a version of any viewer which does not support TLS 1.2 will not be able to do any interactions with cashier or anything that involves money and Second Life.
Also, as I reported in part 1 of this week’s update, the Open Development meeting (also called the open source developer meeting), hosted in-world by Oz Linden, will be moving from Monday to Wednesday from February 17th, the time remains unchanged.