A TPV developer meeting took place on Friday August 1st, 2014. The core items discussed in the meeting are reported below, with timestamps in the relevant paragraphs indicating the point at they are discussed in the video embedded here. My thanks, as always, to North for the video.
SL Viewers
[00:15] There have been no changes to the RC viewer currently in the release channel or to the project viewers. The limited release log-in viewer which is being used for some unspecified A/B testing is still in the wild as well, although this is not available for manual download via the Alternate Viewers wiki page.
The Zipper viewer (for faster installation), which reappeared as a RC viewer on Wednesday July 30th (version 3.7.13.292263), is reported as doing “pretty terribly” crash-wise on older operating systems which may not support the archive format used to zip the viewer skins for faster installation. As a result, the Lab is currently unsure as to whether the project will be continued or not. Until a decision has been made, TPV developers have been advised not to port the code.
The Oculus Rift project viewer was updated to the current viewer code base in week 30. However, it is not currently clear when any updates related to the Oculus DK2 kit will start appearing in the viewer.
As indicated in part 1 of this week’s update, the Experience Keys beta project is going well. There are some issues to be addressed, although these are describing as being “nothing major”, and it’s not clear if they are viewer-related or not.
Unsupported Operating System Versions and Windows 8.0
[38:03] A reminder was issued at the TPV Developer meeting that the Lab will no longer be providing assistance with unsupported operating systems. This includes Windows XP and, once the Library Refresh RC viewer reaches a release status, Max OS X 10.6.
[39:05] It was also reiterated that Windows 8.0 users who are experiencing crash issues with the viewer should consider the free upgrade to Windows 8.1, which has a much lower crash rate, with overall improvement in viewer stability being described as a “really big difference”. A blog post on this subject (and viewer crashes in general) was posted by the Lab towards the end of July 2014.
Group Chat
[03:00] Following-on from the discussion about group chat and the initial testing of updates which took place at the Server Beta meeting on Thursday July 31st, Oz said of the work, “we’re continuing with making back-end changes, there’ll be another roll-out of some experiments shortly and we’ll see how they go. Like I’ve said before, we’re not going to generally announce when those happen, because we don’t want to change the experiment by changing people’s perceptions.”
[10:27] Although at the moment the focus is very much on what can be achieved on the back-end services without the need for changes within the viewer itself, Oz gave notice that the Lab may want to talk to TPVs about possible changes to group functionality and viewer behaviour at some point in the future. One of the questions being asked within the Lab is whether or not the members list needs to be displayed for all groups. Some groups, for example have a “no chat” policy and / or are only for the outward flow of information (e.g. product update groups); so are these a category of group for which the updates of people coming on-line or joining / leaving the session are no actually relevant, and could be eliminated or suppressed, thus reducing the volume of update messages?
NorderNey, Heaven Scent (Flickr)
Cocoa Issues
[03:45] There was a healthy discussion on the subject of the various Mac issues which appear to be Cocoa related.
Some fixes have been supplied to the Lab by TPV developers, although these were reported as making matters a lot worse during the Lab’s testing, so they are unlikely to progress further in their current form. That said, there is a slight lack of conviction within the Lab that there are a lot of high-priority Mac-specific bugs which need urgent attention, and so some agreement appears to be required between LL and TPVs as to what is having an impact and what still needs to be addressed. In terms of the most common issues being experienced by Mac users:
- ALT-cam issues: the view from the Lab is that they’ve fixed one issue in this area (related to the camera zooming strangely when using the ALT key), so if people are still experiencing the problem on recent viewers, it might be symptomatic of another issue
- Alt-clicking while moving mouse moving the camera significantly: a fix for this had been included in the last Maintenance RC (now the current release viewer, 3.7.12.291824, (as MAINT-3171), has been acknowledged as still occurring. however, as this seems to require very exact conditions in order to occur (e.g. clicking twice without moving the mouse at all between clicks), it is not currently viewed as a high priority item
- Keystroke entry lag: this is increasingly seen by the Lab as a system issue, rather than a viewer-specific issue, particularly as it appears to occur more frequently on systems with low memory or (as with their internal testing) on systems where a lot of memory is being used by other applications (e.g. running the viewer alongside of multiple instances of the Safari web browser).
Vivox / Voice
[18:29] It now doesn’t appear as if the Lab will be getting the promised additional Linux update from Vivox which would allow Linux users running older versions of a viewer to update to the most recent Vivox API.
There are also reports that a “significant” number of Firestorm users who are experiencing issues running the more recent Vivox API (4.6.x), although these have proven difficult to reproduce consistently, and the Firestorm team is still looking to file a decent report on the problem.
Finally, there is apparently a voice related race condition in Mac viewers which the Lab is currently investigating.
Other Issues
Texture and Mesh Download Issues
[23:03] There have been some forum threads (see here and here) and at least one JIRA (BUG-6838) relating to texture and mesh download issues and avatars remaining fully or partially grey as a result of SSA failures. Exactly how widespread the issue might be is unclear, although it was noted that those affected all appeared to be in Brazil, and all using the same ISP. However, it also appears that the issue might be related to the use of the AVG firewall, with reports that setting this to interactive mode resolved the problem.
[25:29] The problem of HTTP texture / mesh downloads slowing to a crawl on busy regions has also been identified, and the Lab is working on a fix. The problem is believed to be connection handling starvation server-side, which has previously been the result of viewer making hundreds of open HTTP connections to a server. Changes were made in 2013 to how HTTP connections in relation to mesh downloads were handled, and the Lab is experimenting with a couple of different ways in which things can be further improved. Once they have clarified how to measure the results of any changes, they’ll be engaging with TPVs to help test things.
Persistent Blurred Textures
[34:14] The issue of persistently blurred texture, BUG-6382, which the Lab previously had problems in consistently reproducing, now has a fix which has cleared QA and is likely to be fast-tracked into the next batch of maintenance updates from the Lab, or possibly be included in the next snowstorm build, whichever comes along sooner (it might also find its way into the next update of the library refresh viewer in order for it to get out ASAP).
Again, this issue isn’t the same as the texture thrashing issue which has in the past been reported; the latter is where textures constantly appear to be loading / refreshing as graphics memory is used / cleared / used. Rather, BUG-6382 appears to be limited to viewers running the latest interest list code updates, and leaves certain textures either permanently blurred or greyed out long after texture load activity, as reported by the texture console, appears to have ceased.
Texture Thrashing Issue
However, there has been a report of renewed texture thrashing within one of the viewer test channels (Viewer Bear – see BUG-6786), which appears to be the result of an attempt to fix an older issue (BUG-5970). A further fix is in progress within the Lab. As this is an issue within one of the viewer test channels, only those using that particular build should be experiencing the problem.
Other Items
[16:35] A Firestorm release is due on Sunday August 10th. However, the Lab are considering asking the Firestorm team to delay the release until later the following week. This is to avoid possible conflict with a physical get-together of the Second Life development team (and so presumably will be unavailable to provide advice and assistance to LL’s first line support, should it be required following the FS release).
If I had t’choose between members lists and reliable group chat, there’s no question that I’d take th’latter There should be a checkbox in the group settings t’turn it off, and it should default to OFF for large groups. (Or if not off, at least “throttled” durin’ periods of higher lag.)
LikeLike