Viewer Bug Reports

This issue is to point out the differences in water when comparing a current viewer to pre-PBR (in this case FS 6.6.17). If you're reading this and would also like to point out obvious water changes you've noticed since the implementation of PBR then please do! Images are definitely recommended. Be sure to provide a SLURL for the location the image was taken at and what graphics settings are set to :) Please note that this is not focusing on water reflections

It appears between initial PBR/Atlasaurus and ExtraFPS the behavior of Media on a Prim changed for PBR materials. Previously, MOAP would override the Emissive channel, showing the media, as of ExtraFPS, this no longer appears to be the case, being completely hidden from view now. This has caused one of my products to catastrophically break. Product In Question: https://marketplace.secondlife.com/p/CKC-Cabinet-TV/25515968 (I'd use the "Cabinet TV [PBR, Media, 80s Ads]" version, as you'll know media is playing with the audio) Difference Between Versions: https://imgur.com/a/eI23Jzf

1.) In general, the VRAM usage for textures is still way too high compared to Pre-PBR (V6) viewers. For example a simple scene with 3 avatars visible, a few palm trees and void water in the background, almost 3GB of VRAM was used for textures! 2.) Texture bias calculation is bugged and can lead to constant up- and downscaling of textures: If an over-budget of VRAM is detected, the discard level is increased. When memory is available again, the viewer slowly reduces discard level until an over-budget is detected. Then the cycle starts again and the viewer will never stop and reach a stable state. This is accompanied by the CPU constantly ramping up and spikes in frame times every 4-5 seconds, making it an awful and unpleasant experience. Here is a video of the problem when observing the texture console: https://gyazo.com/57cb4d19a9cd24eea22a4a8f0e52fcd2 3.) Bug #2 in texture bias calculation (in LLViewerTexture::updateClass()): While downscaling textures by increasing the discard bias, the viewer waits 1 second after a downscaling before another possible downscaling happens. However, if memory isn't low anymore, the decrement of discard bias and upscaling of textures happens every frame.

Using the latest SL Viewer, data usage seems about double from earlier versions - specifically the latest full release of FS v7.1.11 Visiting a single store and caming around for a short time (app 20 minutes) uses 2.63 GB according to Windows. Visiting the same store on FS (7.1.11) and caming around in similar fashion uses 1.37 GB of data. Both viewers are on MID-ish settings, no mirrors, some reflections, DD of around 64M. Yesterday I visited about 10 stores and used almost 10 GB of data. At this rate, I will start to run into data limit issues with my ISP pretty quickly on the new viewers (the issue also seems to affect FS recent beta versions)

I saw this in one or two spots recently, but no real clean bug report, so I am writing this. Note, that I've done this often in my 10 years of working in SL, and I don't recall this always happening but behavior seems to have changed recently. When you link 2 things together, and then attempt to corner stretch a single object in the linkset, it acts like "Stretch Both Sides" is enabled, when it's not. It only does this on corner stretch (not one-dimensional stretches) and only when part of a linkset. You can see me demonstrate this here: https://gyazo.com/b5c2a54823ccf28dcbc3fc127a847035 Steps to reproduce: Create two cubes. Link the two cubes. Right click Edit the linkset. Click Edit Linked Click the child prim Click Stretch Attempt to corner stretch the prim, observe bug occurs. Attempt to one-dimension stretch the prim, observe bug does not occur. Unlink the cube, and then repeat steps 6, 7, and 8. For step 7 and 8, the bug does not occur in either case when unlinked.

Cleaning out my block list I noticed a couple entries that appear to point to avatars, but don't display proper names: instead the entries read hindare20completeparam1l1n02.lindaprivat… and lindar011487302797692204159.lindaprivat… If I right-click them I get options to unblock, but trying to open the profile fails. (See screenshot.) I'd be curious what these entries represent. Deleted accounts? Banned accounts? Corrupted data? Something else?

I either get a crash, or a hang or no issue when quitting the viewer. I was told it was a problem with the upstream viewer from Linden Labs by Megapahit support staff and that I might want to report it. Thank you.

This is not really a bug report, but more of an observation. I noticed the Mac viewer sees the framerate commonly divided by about 3 when activating shadows (sun/moon), for instance from 75 to 25, while a PC with an AMD integrated gpu, with the same settings and at the same place sees a more reasonable drop from 65 to 40 fps. Its worth noting that the amd igpu utilisation is always maxed out (it's about twice weaker compared to the M2 Pro gpu), shadows on or off, while the Mac gpu sees its gpu utilization dropping significantly when activating shadows, indicating that shadows are limiting its efficiency in some way. The cause can be driver related of course, but i thought i would leave this observation in case you find a way to optimise this. Also i don't think this is due to "foreverfps", it must have been something going on for some time, but i really noticed this only today. I also noticed this on a 3rd party arm compiled viewer, so this is not a Rosetta translation issue.

Blinn-Phong textures with alpha channels appear excessively transparent in the PBR client, with thinner alpha edges. Adjusting the exposure to the lowest setting or changing environment settings doesn't help. This issue affects many products, and the only solutions are to double the layer or make textures more opaque and have two versions of the same product included with the purchase. However, this causes inconsistency for users who don't use PBR, as they will see the items differently than PBR users.

I just bought a full perm baseball cap, well 2 of them and unable to save their textures to my computer, takes like 30 seconds for a dialog to come up Unable to download. I tried with like 4 different textures, even did a relog.