The following notes are taken from the TPV Developer meeting of Friday February 14th, a video of which is included below. My thanks, as ever, to North for the latter. Timestamps relative to the recording are given in braces both at the start of each section and within the text where appropriate.
Release Viewer Update
[01:20] The Facebook Hotfix RC (version 22.214.171.1246557) released on February 12th was promoted to the de facto release viewer on Friday February 14th. Based on the Fitted Mesh viewer code, this viewer has a single fix for a problem in the in-viewer web browser that made it impossible to login correctly into Facebook. The rapid promotion was made because the issue was seen as obnoxious by those people who have connected their SL and Facebook accounts, and it was felt those who do may want to post Valentines-related messages and images to their Facebook accounts.
However, it has suffered from the number of RCs currently in the queue, or as poor Monty put it, getting “stuck behind everyone”. This viewer has one of the lowest crash rates on record as an RC, and given this, the expectation is that it will be promoted to release status “pretty soon”.
One of the major issues Monty faced with the viewer-side updates was directly related to mesh, and thread race conditions, and he admits that not all of these have been resolved. This is partially due to some of them being infrastructure-related Heisenbugs, which are time and labour-intensive to resolve. However, they shouldn’t impact the stability of the updates made to date.
Remaining RC viewers
[03:54] The three remaining RC viewers – the Maintenance RC (126.96.36.1995499), Interest List RC (188.8.131.525213) and Google Breakpad RC (184.108.40.2065686) are in the process of being rebuilt to the 3.7.1 release code, so updated versions should be appearing in the release channel in the next week (ish).
- It is unclear whether the Merchant Outbox project viewer (220.127.116.114731) will move forward or pulled back to have some more work done on it, and it is unlikely to move towards viewer release “any time soon”
- The Sunshine / AIS v3 project viewer was rebuilt to the 3.7.1 code (Fitted Mesh and the Facebook Hotfix), with a new version (18.104.22.1686565) appearing on February 14th. It is anticipated that this viewer will move forward to a release candidate status fairly quickly now that the Facebook Hotfix has been promoted, reducing the number of viewer RC cohorts currently in the release channel.
[05:16 and 14:50] A surprising piece of news passed-on in the meeting was that the AIS v3 server-side code has been deployed across all channels on the Main grid. This initially caused some confusion during the meeting , as there has been no mention of this in any server-side release notes. Nyx Linden queried the situation with the ops tem and received a confirmation that the new AIS capabilities had been deployed to Agni [21:40, via text], but are currently disabled [22:29 via text]. It would seem likely that the capabilities will be enabled once the Sunshine / AIS v3 project viewer moves to release candidate status.
[07:35] There have been a number of issues with regards to voice in SL, particularly of late. As noted in my week 6 report, there was some recent back-end work carried out which should improve things for those using viewers running with the most recent versions of the voice SDK (SLvoice.exe).
Discussing the matter at the TPV Developer meeting, Oz Linden revealed that Vivox had reached out to the Lab to assist with issues being experienced, and as a result of this underlined the issues with viewers using older versions of the SDK (and which will not see any real improvements to their voice performance as a result). Vivox have requested that TPVs provide details on any older versions of the SDK they are running, and details of specific issues they are encountering, as well as offering encouragement to update.
As a new version of the SDK is due to be released in the near future (hopefully within a couple of weeks), it may prove to be an opportunity for TPVs to update, given it has a number of audio quality fixes and Vivox have offered to assist in dealing with issues being experienced with voice in SL.
In addition to this, Oz is looking to work with Vivox to try to get any new versions of the SDK used by the viewer made available to TPVs at the same time it is made available to the Lab, thus eliminating the need for TPVs having to wait for LL to QA and integrate the package into the LL code prior to being able to merge it into their own code, allowing them to test new SDK releases in parallel with the Lab. These will hopefully include 64-bit binaries of the SDK as well a 32-bit versions.
Group Ban Lists
[48:23] It had been hoped that Baker Linden might be providing an update on the overall status of his group ban list work. However, this was unfortunately not the case, although Oz provided a small update on things, stating that he has been able to sit down with Baker to review the updated viewer code, which is now with QA. Hopefully this means it will be appearing in at least a project viewer in the near future.
[39:02] Mac users running versions of the viewer with Cocoa updates are continuing to report significant issues (BUG-1785), including frequent viewer crashes, and it is a little unclear as to what is being done to address matters. Monty Linden, who uses a low-end Mac system for testing purposes, can anecdotally agree that there are Cocoa-related issues, and Oz has requested TPVs contact him via e-mail with specifics, so that matters can be carried forward.
Mesh Upload Issue
[45:50] There is an issue related to Fitted Mesh and the viewer’s mesh uploader, with the latter having the wrong joint matrix size and missing bounds check in the mesh upload preview. There is a possibility that the issue may be down to an incorrect array size, although this is not (as yet) clear. A JIRA has been raised on the issue and the Lab is aware of it, and further information is being passed to Oz on the matter.
LookAt Name Tags
[23:07 – 36:00] Almost a quarter of the meeting was taken up by a discussion of some viewers displaying avatar names above their LookAt crosshairs, with accusations of “drama causing”, and/or “an invasion of privacy”. The discussion was somewhat reminiscent of the arguments around LookAt beacons in general which occurred a few years ago.
The debate led to some interesting points being made about the validity or otherwise of making viewer capabilities intended for debugging purposes generally available as viewer features, and whether or not Linden Lab should police every single option within the viewer in terms of whether its broader use should be allowed. I’ll say no more here, as the issue falls somewhat outside of a project report, but will leave it to you to listen to the debate / discussion within the video.