Running slightly behind in weekly updates due to the small matter of RL house decorating. I long for the day when I can take a colour swatch and simply drag my choice from the swatch onto the wall and see the wall instantly painted…
Server Deployments: week 17
The Main (SLS) channel and the LeTigre and BlueSteel RC channels all have no scheduled deployments for the week.
On Wednesday April 23rd the Magnum RC received the same server maintenance package as deployed to it in week 16, but with an additional fix for BUG-5763 (“AGENT_MOUSELOOK flag is often incorrect when llGetAgentInfo() is called within control() event”). This bug was a regression in the original update, and which affects certain guns that only allow some actions when they think the user is in Mouselook and certain vehicle that change control behaviour when in Mouselook. Early indications are that the fix has been successful.
SL Viewer Updates
On Monday April 21st, the VoiceMO release candidate viewer, version 3.7.6.289164 ((download page, release notes) was promoted to the de facto release viewer. This viewer pulls in Vivox 4.6.x libraries instead of 4.5 for improved stability and to address Mac Mavericks issues as well as fixes for accurately detecting Merchant status and improves recovery for Merchant Outbox errors. It should also removes the Windows crash reporter pop-up notification.
The Maintenance RC updated on Tuesday April 22nd to version 3.7.7.289405 (download and release notes). This RC includes over 50 MAINT category fixes broadly covering:
Multiple fixes to Mac viewer
GPU table updates
Fixes in Recent tab, Chat, LSL editor, land management, etc.
Help system update
Crash fixes
Performance improvements
The interest list RC updated on Wednesday April 23rd to version 3.7.7.289461 (download and release notes). I’ve still not found the time to have a play with this viewer.
The sunshine RC should update as well, most likely before the end of the week. Please refer to my current viewer releases page for updates on SL viewer versions and TPVs listed by LL.
STORM-1831: LSL Syntax Updates
STORM-1831 is the work mainly undertaken by Ima Mechanic (with assistance from Oz Linden and Cinder Roxley) to improve syntax highlighting in the viewer’s LSL editor by allowing the viewer to obtain the information required for syntax highlighting directly from the simulator the viewer is connected to. This should eliminate issues with the current manually updated files used to manage syntax highlighting falling out-of-synch with new LSL syntax as new functions and parameters, etc., are added. Folded-in to this work should also be a change to the source code text allowance in the viewer’s LSL editor, increasing it from the current 65,000 characters to around 256,000.
The server-side support for this commenced deployment in week 2, but the viewer work has been subject to delay for a variety of reasons (not the least of which has been the number of RC and project viewers in the release and various project channels). However, it is hoped that a project viewer with this work (and possibly other STORM contributions) will be appearing either this week or in week 18, and Oz hopes to see some testing carried out using it once it does so.
Group Chat Optimisation
Little to report on this, other than the Lab is “working out some kinks”. Further details on the status of Agni testing may be available at the Server Beta meeting.
There were no deployments to the Main (SLS) channel or the BlueSteel and LeTigre channels during week 16.
On Wednesday April 16th, the Magnum RC received a new server maintenance package, which included a fix for BUG-5533 (“llTeleportAgent() and llTeleportAgentGlobalCoords() can break any script in any attached object that contains a change event.”).
Commenting on the latter at the Server Beta User Group meeting on Thursday April 17th, Maestro Linden said:
Unfortunately, after the roll we discovered that there was a regression in the Magnum update, which was BUG-5763 (“AGENT_MOUSELOOK flag is often incorrect when llGetAgentInfo() is called within control() event”), which apparently affects certain guns (which only allow some actions when they think you’re in Mouselook) and certain vehicles (which change control behaviour when they think you’re in Mouselook).
The good news is that Kelly [Linden] came up with a fix for it yesterday, which is out on Aditi now. I gave it a whirl, and it looks like this bug is fixed. But if you have content that was broken by that bug, I’d encourage you to test it out on Aditi, in case there are any additional problems.
Those wishing to test the fix can do so on the Aditi regions Ahern, GC Test 10, and Tehama. GC Test 10 is probably the most convenient to test on, as it mostly allows anybody to build and run scripts.
Week 17 Deployments
While the final details of deployments for the week commencing Monday 21st April will not be determined until the start of that week, it currently looks as though there will again be no deployments to with the Main (SLS) channel or the BlueSteel and LeTigre RCs, while Magnum will gain the BUG-5763 fix.
Commenting on the lack of high-profile server updates of late, Maestro pointed to the fact that the Lab has been engaged in a series of “invisible updates” recently, notably infrastructure improvements.
Group Chat Update
There were no further tests on Simon Linden’s group chat work, and Maestro indicated that testing on Agni may commence in week 17. An idea initially discussed for testing these optimisations on the main grid had been to use a single large group (the Firestorm Support group, due to both its size and frequency of use). However, commenting on the work, Maestro Linden said, “since the group chat changes are in the backend service, this would mean that … around 1/16 of groups would be on the group chat stuff.”
Other Items
HTTP-in Failures
Some people are noticing an uptick in issues relating to in-world scripts acting as HTTP servers (notably with HTTP-in functions). There has been a known bug with these (non-public BUG-2564) wherein all http-in URLs and all capability URLs for connected users are dropped simultaneously, all the connected users get logged out, and HTTP-in scripts cannot be contacted. However, some of the issues people are experiencing appear to be occurring since the most recent HTTP updates were made. Lucia Nightfire describes the problems as, “random URL loss and instability is common esp after the http changes.” she goes on to note that she has had to “change http protocol and add heartbeats to some apps since the HTTP changes.” She further goes on:
Sometimes after I crash then relog into a region, its like my caps are reset and re-evaled and in-turn all my HTTP devices all of a sudden cannot request URLs until the caps are reset or it is [a] repeat URL request failure. Long story short, if your viewer crashes, don’t log into your home region if you have servers there, or you risk interruption.
Maestro’s thinking on the matter is that it may be linked to a server crash – particularly given Lucia confirms the problem does occur as a result of a region crash – as there is nothing specific to a viewer crash which should upset things like HTTP-in functions. Further testing has been suggested to see if a precise cause can be identified.
The SL Share 2 project viewer, which includes the new options to upload text and snaps to Twitter and snaps to Flickr, complete with post-processing filter capabilities, and with the updated Facebook snapshot upload tab (also with the filters), was updated on Tuesday April 15th to version 3.7.6.288806.
Other Items
AIS v3 Inventory Outfit Links Issue
For those who use the Outfits folder in inventory, a useful option is that of being able to create links to no Copy items, allowing them to be “shared” among multiple outfits. This capability was originally introduced in Viewer 2, and has been popular among a number of TPVs.
While testing the new AIS v3 code, Henri Beauchamp noticed that when creating a new link in inventory in an AIS v3 region and when using an AISv3 compatible viewer, the AIS server reply is not properly parsed, resulting in a failure to sync the viewer-side inventory (and its cached list) with the server-side inventory.
He has filed a JIRA (SUN-129) relating to the issue, although it is not clear if the issue is unique to Cool VL viewer, which uses a different algorithm to create the COF links, or whether it may affect other viewers still using links.
MAINT-535 Viewer URI Name Space
Viewer URI Name Space is a means of specifying URIs in a format the SL viewer can understand and use. For example, “secondlife///app/teleport” can be used to display a teleport link in local chat, IM or from a browser such that when the link is clicked, the user is automatically teleported to the destination. For example: “secondlife:///app/teleport/ThisPlaceHere/71/6/22” would display a link “Teleport to ThisPlaceHere” in chat / IM which when clicked, would teleport the user directly the destination.
However, because the capability can include an innocuous label, rather than displaying the actual SLurl, there have been concerns that the capability could be abused. For example “[secondlife:///app/teleport/GoHereAndGetBanned/3/4/56 CLICK FOR FREE STUFF!]” would be displayed in chat or IM as “CLICK FOR FREE STUFF!”
Because of this risk, Linden lab updated things such that the use of Viewer URI Name Space causes the official viewer to display a warning:
A couple of side effects of this are that all SLurls which use Viewer URI Name Space can be impacted, and other uses of the capability can also be affected as well, as is the case with BUG-5702.
To address the issue with teleports, the Lab has introduced Maint-535 (currently in the Maintenance RC viewer, version 3.7.6.288799. This works by opening a dialogue box which displays the actual destination, together with button for the user to either teleport or ignore the offer, as shown in the example below.
MAINT-535 causes teleport links which use Viewer URI Name Space to generate a dialogue box displaying the destination, regardless of any label used to disguise it (click for full size)
Do note that all of this is only relevant to SLurl which use Viewer URI Name Space, Slurl from the map, landmarks, etc., do not see any change in behaviour.
Some concern was raised at the Simulator User Group meeting on Tuesday April 15th that because the dialogue box is displayed for all teleport offers using Viewer URI Name Space, even those that display the destination (e.g. “Teleport to ThisPlaceHere”, and that it might cause annoyance among people as a result of the need to confirm a teleport.
MAINT-535 also doesn’t address issues such as BUG-5702, where the “untrusted browser” message continues to be displayed.
A TPV developer meeting took place on Friday April 11th. 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 latter.
Release Candidate Viewers Status
[00:34] The VoiceMO RC (version 3.7.6.288881) combines the former SL Voice RC (3.7.5.288516) and the former Merchant Outbox RC (3.7.5.288408), both of which have now been withdrawn from the release channel. Both of these RCs were performing well, and it is hoped that by combining them, they’ll both have an accelerated path through to formal release status.
The Lab is keen to see the Vivox updates for voice gain wider traction in viewers, as these not only improve people’s voice experience, they also lighten the load on the back-end servers.
The Interest List RC updated to version 3.7.6.288879 and the Sunshine / AIS v3 RC updated to 3.7.6.288822, both dated April 10th, and appearing on the wiki page on April 11th.
These, together with the Maintenance RC (version 3.7.6.288799), released on Wednesday April 9th, are now the four remaining release candidate viewers in the release channel. As these are all recent updates, data is still being gathered on all four, so it is unlikely that any of them will be promoted to the de facto release viewer in week 16 (week commencing Monday April 14th).
Google Breakpad Reporting and Issue
[01:40] The Google Breakpad changes mean that all viewers with the updated Breakpad code have the marker files for crash reporting created much sooner in the initialisation process and deleted much later in the viewer shutdown process. This means that more crash stats are being collected, but it also means that the crash rate measurement has gone up slightly as a result.
[03;30] Whirly Fizzle has identified an issue occurring with both the Maintenance RC and the current viewer release (3.7.5.266464) which appears to be linked to the Google Breakpad updates. BUG5707, “While logged in, crash reporter pops up, sends in a crash report but the viewer does not crash”. Whirly describes the issue in part as:
The crash reporter window does not just sometimes flash up briefly at viewer launch or viewer close, the crash reporter stays on-screen and appears to be actually sending in crash reports – I see the “Sending to server…..” window, which stays onscreen for some time.
Often I have more than one “Sending to server….” window open – often as many as 3 of them will all open at the same time.
This is happening on viewer launch, on viewer close and even while logged in.
When it occurs while logged in, the viewer does not crash, I am still logged in and the viewer appears functional.
Commenting on the situation, Oz Linden said, “Yeah, we’ve got people looking at some of the Google Breakpad crash reporting issues … I’m not really sure what that’s about … we’ll see if that can get some refinement.”
Oculus Rift
An update to the Oculus Rift project viewer, currently on closed beta, is expected in the next week. The lab is still accepting applications from those who have the hardware to join the closed beta the requirement being that applicants actually have the Oculus hardware. Applications should be addressed to sl_oculus_beta@lindenlab.com.
Further updates to this viewer are expected, particularly once the Lab received the SDK2 headsets, although these are not anticipated to be arriving for another couple of months. The viewer most likely won’t be opened-up publicly until after these latter updates have been made, so TPVs will not be able to integrate the code for some time to come.
SL Share
[05:26] Merov Linden was on-hand to talk about the recent changes to the SL Share capability in the SL viewer – notably the updates which allow people to post pictures to Flickr, post text and images to Twitter and once again post images to Facebook. It is likely that this viewer will be progressing to release candidate status in week 16.
As I’ve previously noted, an important thing to remember about SL Share is the all of the capabilities are opt-in. No-one is forced to use them, and the Facebook option in particular has nothing to do with trying to “push” SL users to Facebook, and while Facebook may not allow people to have accounts in their avatar names, there are a lot of people who are happy using it to connect their SL account with their real life account on Facebook.
One issue that did cause a problem was the inclusion of SLurls with photos being uploaded to Facebook, which caused the image upload capability to be blocked by Facebook. The SL Share updates have removed the “include location” option in the photo upload tab of the Facebook floater, and as a result, Facebook have removed the block once more.
The SL Share photo upload panel as it was prior to the change (l) and as it appears in the SL Share project viewer and some RC viewers since the change (r)
The issue here was not so much that SLurls were being included in uploaded photos, but the fact that they were being added automatically as a part of the automated processing of snaps during the upload process. Apparently, a requirement in using the Facebook API is that all text uploaded or appended to images must be manually entered by the user …
As well as restoring the upload capability, Facebook also restored photos with SLurl include which had been uploaded prior to the block coming into effect, again apparently because of the SLurls being automatically added.
One of the most interesting parts of the new capabilities is that of the preset filters within the upload panels for Facebook, Twitter and Flickr, and the ability for people to create their own filters.
When SL Share 2 appeared, questions were asked on various blogs on whether the filter capabilities would be added to the viewer’s snapshot floater, so that they could be used with the snapshot options there – most notably with the profile feeds option.
A preview example of a snapshot using one of the SL Share built-in filters
I took the opportunity at the TPV developer meeting to ask, and Oz Linden replied:
The SL Share project doesn’t include putting the filters in to the snapshot floater, but Niran [NiranV Dean, creator of the Black Dragon viewer] is working on that. So we may get that as another open-source contribution.
Merov then added:
It’s not overly difficult to do that. Actually, I did it for a demo once, so it’s pretty easy to add. I didn’t do it because the snapshot UI on the official viewer has plenty of … how can I put this nicely? …. Oddities, let’s say … so if I really wanted to do it, I’d have to re-do everything …. so I went to the designers and said, “You know, we should really re-do all this,” and they said, “Yes, but not now.” so I’m not working on that just now. but yes, it’s possible … so if you want to put it into your own snapshot floater, I won’t be offended!”
One of the aspects of the SL Share 2 capability that didn’t get to be in this initial update is a filter editor capability such that the filters can be more easily combined to create custom filters. Oz invited any TPV developers who might like to take this on to do so and to consider contributing it back the Lab should they do so.
The authentication aspect of the SL Share capabilities is all handled by the Lab’s back-end to the capability, so that the viewer-side code can be taken by TPVs and amended without any risk of users’ private information being leaked or tracked.
Webkit Woes
Monty Linden
[15:48] Webkit is a third-party library used within the viewer for a number of tasks. For example, it powers the built-in web browser, and is used to display profiles (unless you’re using a viewer supporting legacy profiles). It is also used with like Media on a Prim (MOAP) and many in-world televisions.
There have been an increasing number of issues with webkit. The libraries used within SL are out-of-date, for example, something which has caused the Lab and TPVs a considerable amount of pain (see BUG-4763 and FIRE-12642, and FIRE-11057).
Monty Linden has been poking at the problems, and gave a further update on his work:
I’m doing a bunch of damage, and then trying to move forward, just to get to 4.7.4 first, and then the 4.8 a little bit later on. I’m just about at the end of that … my goal is to kill the current third-party lqtwebkit library repo and I’ve created a new one that’s a little healthier and working a little bit better.
It’s not clear at the moment when Monty’s work will appear in a release stream. A further problem here is that the qtwebkit (on which lqtwebkit is based) has been deprecated by QT, so the Lab is faced with a decision s to what to do going forward. One option may be to go with CEF, but which direction the Lab will take has yet to be decided.
The Firestorm team have reported they have re-worked webkit for themselves, updating to version 5..2.1 for Windows and are looking to do the same for Linux and Mac. This work appears to have fixed issues with Media on a Prim and it is hoped it may also resolved the YouTube issue of videos failing to play on in-world TVs, etc.
Firestorm are going to liaise with Monty with regards to the work they’ve done. currently, the Windows updates are still subject to internal testing by Firestorm, although they plan to make the work visible to all once they are confident enough that it behaves as expected.
Mac / Cocoa Update
[22:45] There has been some progress in addressing the Mac / Cocoa issues, with some fixes now in the Maintenance RC release, including:
MAINT-3135 Cocoa Viewer: Mac: Maximizing the viewer leaves garbage on the screen
MAINT-3288 Mac: Fullscreen mode issue in Viewer 3.6.7 (281793)
MAINT-3642 Mac viewer can no longer export textures to TGA format
MAINT-3674 Mac: “Hide NewApplication” under Second Life menu, should be “Hide Second Life”
Referring to Cocoa issues in particular, Oz again confirmed that “all of them” are on the Lab’s to-do list, but not all of them are getting done as yet. In particular, there is no news on the long-standing ALT-cam bug.
Other Items
We’re all aware that there are options within the viewer to run more than one instance (so if your computer doesn’t grind to a halt, you can run two versions of the official viewer or two versions of Firestorm, etc). However, what some people may not be aware of is the fact that while the viewer can allow multiple instances, this is not actually a supported feature. It should also be remembered that because of the way voice is handled, it is currently not always possible to run two instances of a viewer (or two different viewers at the same time) with voice enabled without encountering significant issues and potential lock-ups (although Singularity has a fix for this).
What also may not be realised is that because some of the viewers use different mechanisms to detect how many instances they are running, it is possible to encounter conflicts and issues when running, for example, the SL viewer and the Firestorm viewer simultaneously on a computer.
So if you are running multiple instances of the same viewer or multiple viewers on the same computer, and start experiencing unexpected issues, try shutting all but one instance / viewer down and seeing if the problems go away before seeking support assistance.
As noted in part 1 of this report, week 16 should see a server maintenance update deployed which includes a fix for BUG-5533, “llTeleportAgent() and llTeleportAgentGlobalCoords() can break any script in any attached object that contains a change event.”
Commenting on this at the Server Beta User Group meeting on Thursday April 10th, Maestro Linden said:
Kelly found some race conditions which likely lead to the breakage, and the fix appears to work. Right now it’s on Aditi, in the DRTSIM-251 channel. Ahern is on that channel, though I found out that almost all of Ahern’s parcels have scripts disabled. However, “Tehama” is also on that channel, and does allow scripts on some of its parcels. It’s public access, so that’s a good place to check the fix out if you’re interested.
Server Beta Meeting, Thursday April 10th
SL Viewer
On Wednesday April 9th, the Lab release the VoiceMO release candidate viewer, version 3.7.6.288881 (release notes), into the viewer release channel. This RC combines the updates found in the Voice RC viewer (3.7.5.288516) and the Merchant Outbox RC viewer (3.7.5.288408), both of which at present also remain in the release channel for the time being.
Also on Wednesday April 9th, the Lab issued a new Maintenance viewer RC, version 3.7.6.288799, which includes some 54 MAINT fixes from the Lab – see the release notes for details. However, this does include a fix for the Male Avatar Chest Rendering Broken / Mesh Clothes Breaking on Male Avatars (BUG-5537) issue I covered in week 13. The fix for this is given as “MAINT-3896 Male Avatar chest rendering broken – chest shows bigger for onlookers, than to oneself – effects attachments and all worn mesh types”. This update also includes a change to remove the “Start Second LIfe now?” dialogue which is displayed after installing the viewer, and which resulted in the viewer being started with system Admin privileges on Windows systems.
Commenting briefly on things viewer, Simon Linden indicated that he’s also fixed the bug where terrain textures aren’t updated if you change the heights or textures, although I gather this has yet to reach a viewer.
Group Chat
A further Group Chat test was run on Aditi. “I’m looking into the overhead for group chat on maintaining that list of people in the chat,” Simon explained as the test kicked-off. “I think it becomes heavy load in a group with a lot of people on-line.” The test proceeded along similar lines to those carried out previously.
Group Bans
Baker Linden reports that most of the viewer-side bugs related to group bans have been fixed. however, Caleb Linden found an ugly bug in the back-end code whereby only the first name on a list of people being invited into a group would be checked to see if that person had been previously banned (and thus prevented from joining); anyone else further down the list would had been previously banned would be allowed to re-join on receiving the invitation.
A fix has yet to be written, but Baker doesn’t see it as a hard task to do, “My fix will check each one and if you’re banned, it will gracefully skip the banned agent’s invite,” he said when describing how things will work. Other Items
Aditi Log-in Issue / Inventory Update Issue
The script which should synchronise people’s passwords and inventories between Agni (the main grid) and Aditi (the beta grid) is not functioning correctly (see BUG-5563). As I’ve previously reported, this means that following a password change, people are finding they must continue to use their old password to access Aditi and are not seeing their inventory update. Commenting on the issue, Maestro Linden said, “We’re not sure why it’s not working – the script which is supposed to synchronize the accounts says that it updates them on Aditi.” So at the moment any fix is dependent on the Linden carrying out further checks to ascertain what is going on.
Feature Request: Hide All Objects Outside Parcel
A request for a parcel control feature which, when set, would cause the viewer to ignore and not render all objects outside of a parcel has been put forward (see BUG-5671) and is drawing strong debate.
On the one hand, people feel this could greatly enhance performance when entering regions that are massively loaded with objects and textures (because the viewer would simply ignore everything outside of the parcel where the option is enabled).
On the other hand, some people feel that such an option could negatively impact the feeling of SL as a place (imagine teleporting around regions and all you see are the objects on the parcel you’re in, the rest is just terrain).
One suggestion put forward is that rather than making this a parcel option (and thus forced on anyone entering a parcel where it is enabled), such a capability is, if possible, made a viewer-side option, so that people get a more flexible choice as to what is rendered / what they see. I have to say that this potentially offers the most attractive option were such a capability to be taken-up by LL.
It’s pretty simple again this week: there are no server deployments scheduled for either the Main (SLS) channel (which remains on maintenance project 14.03.12.288004) or the RCs (which collectively remain on 14.03.28.288552). As always, please refer to the server deployment thread in the forums for the latest news and updates / issues.
SL Viewer
The de facto release viewer updated on Monday April 7th, with the promotion of the Google Breakpad RC, version 3.7.5.288464 (release notes).
As a result of this promotion, the StatTest viewer (formerly version 3.7.5.288371), which was never intended for promotion as a release viewer but issued as a means of assisting with bug-fixing the Google Breakpad code, has been removed from the viewer release channel.
Group Ban Lists
“I’ll be working with Maestro this week to try to get the group ban services and back-end stuff grid-wide on Aditi,” Baker Linden informed the Simulator User Group meeting on Tuesday April 8th. He went on: “I have a handful of smallish bugs to finish up, and then I’ll look at my options for development viewers and such, so group ban will be grid-wise soon [on Aditi] (I hope, depending on the amount of releases in the pipe). And then after we see there’s no big issues, we’ll get that pushed out to Agni.”
Other Items
Aditi Log-in Issue / Inventory Update Issue
As reported in part 2 of the my week 14 updates, there has been an issue in getting passwords and inventory to correctly sync on the Aditi beta grid following a password change. Normally, a script is run on daily basis during periods of relatively low SL use (around midnight-02:00 SLT) which should synchronise a user’s password and inventory between Agni and Aditi. However, several people had noted that their Aditi passwords / inventories were not updating despite several days passing after making password change (see BUG-5563).
It had been thought this issue had been dealt with in week 14; however, a forum thread notes it is still causing problems for some people.
Transaction History Oopsie
There was a brief issue with Transaction History pages on Tuesday April 8th which caused some consternation when it happened, although it was quickly rectified.
The problem first came to light when merchants noticed that their Transaction History page was no longer showing totals or options to download the history in anything other than .CSV format (the previous options had been .XML or .XLS). Further issues were noticed as time went on, as noted in a Commerce forum thread and also in a JIRA (BUG-5664).
The root cause of the problem appears to have been the URL for the familiar Transaction History page being swapped for a new page. The concerns this raised were sufficient for the URL to be reverted back to the original a little over an hour after the problem was first noticed, allowing people to once more access the familiar Transaction history page. Whether the change of URL is indicative of an upcoming change that is in preparation, or simply a mistake on the part of someone at the Lab, is unclear.
Inara Pey: Living in a Modemworld 