Avatar attachments STILL failing to appear in busy regions
in progress
animats Resident
This old bug is still present, despite claims to the contrary. See the attached picture.
Nude and broken avatars are on display for many minutes.
Log In
Chorazin Allen
Maestro Linden I've been looking into zombified HUD attachments with Kokua and am coming to a similar conclusion to Henri Beauchamp in that the simulator isn't sending all of the replies required for an ObjectPropertiesFamily request.
My repro is fairly reliable and involves using several bulky huds (30+ linked objects) notably Tiny Empires and Tiny Empires 3000 along with the Maitreya and Lelutka HUDs. I can get a failure most times when teleporting from Bexington to Gormthong. Although I'm testing with Kokua and lots of additional logging, I can get a failure using the standard 7.2.1 viewer too.
The failure manifests as Edit not being enabled in the menu for an affected HUD when it is right-clicked. Attempting to click the HUD may succeed but it will become unresponsive once an invalid child object is clicked.
What's happened under the covers is that ObjectProperties replies appear to be limited to a maximum of five objects (meaning seven messages will be needed for the whole linkset) and not all of the replies have been received resulting in the linkset as a whole being considered invalid for editing.
This doesn't seem to be a region arrival data burst thing. If a different HUD is selected and then the affected HUD is reselected ObjectProperties gets requested again and will consistently return the same incomplete response.
The HUD will (probably) work again after a teleport into a different region.
This seems to suggest that the simulator has an incorrect record of the linkset and only believes that it consists of a subset of its full complement of child prims. This appears to be confirmed by using object inspection which only reports a few child items rather than the full 30+.
I'm happy to co-operate with doing tests so that server logs can be grabbed. Contact me inworld if you'd like to arrange something.
--Chorazin
animats Resident
I have not seen this in Firestorm 7.1.13. Even in the busiest clubs. It was bad prior to Firestorm 7.1.13, but seems to be fixed. Anyone else seeing this problem on a current viewer?
Maestro Linden
in progress
animats Resident
This looks like a problem in Firestorm 7.1.9. I've been going back and forth between 7.1.9 and 7.1.11 at Warehouse 21, with 68 avatars present.
7.1.9 is prone to avatars stuck in cloud mode and avatar attachment loading failiures. In 7.1.9, four minutes after login, most avatars are still in pink cloud mode, a few are missing attachments, and some are fully loaded.
7.1.11 does not seem to have either problem. In 7.1.11, all avatars load successfully within 90 seconds.
(All on LInux, 6 CPUs, NVidia 3070)
Maestro Linden
animats Resident: Given your comment about this being a bug in Firestorm 7.1.9, can we close this issue as fixed in Firestorm 7.1.11?
a
akleshchev Resident
Maestro Linden There is https://github.com/secondlife/viewer/issues/3037 which hasn't been released and beyond that some meshes have low quality lods which may look like there is nothing because your shirt is a single triangle under the skin. 'single triangle' case probably can be handled better, but I doubt there is a perfect solution for that one.
Maestro Linden
Ok, I agree that viewer#3037 could lead to the symptoms reported here. I'll link to that issue as well.
Maestro Linden
needs info
Maestro Linden
animats Resident could you visit one of the big regions again, and attach a viewer log from the session if the missing-attachments issue reproduces? I can't immediately reproduce this myself - other agents' attachments appear to load properly when I visit 'Exhale Dance Club' with an empty cache.
I agree Monty's comment that it's suspicious for the "lost packets" stat to be 2 when in-world objects are failing to appear.
Maestro Linden
Hi Animats, thanks for the report. That 200mbps bandwidth usage indeed seems suspicious and possibly responsible for some level of packet loss in the UDP messages announcing avatar appearance. Out of curiosity, what is your viewer bandwidth slider set to? Do you have an idea if the majority of incoming data is asset fetches from the CDN?
Monty Brandenberg
> Bandwidth: 1500 kbit/s
Which will apply to UDP traffic.
I would like to see the viewer log from such a session. Just don't use this tool for that. :(
Add'l: Also want to verify that that packet loss report was accurate for the target region. Only 2 lost UDP packets? animats Resident