Cool VL Viewer forum

Here is a new release of the Cool VL Viewer, available now from http://sldev.free.fr/:

This is a maintenance release and a recommended update.

Cool VL Viewer v1.32.0.17 (stable branch):

New in this release:

• Fixed a bug introduced in last release and which caused voice speakers to be wrongly flagged with a muted icon when not speaking in the "Active speakers" floater.
• Fixed the "Make outfit..." button status staying grey after login in the Appearance floater when using AISv3 for inventory fetches in SL. See this message for details
• Properly take into account the chat bar visibility when positioning the media HUD on zoomed media faces, so that the said HUD does not get hidden behind the chat bar.
• Implemented support for loading the media in the connected sim(s) when switching on media on a prim in the settings while it was off before entering the sim(s): this algorithm does not sacrifice the savings in memory usage, network traffic and fps rates when media is turned off, but avoids to have to relog or TP away and back in after switching media on. Now also properly hide the media HUD for faces with media after turning media support off.
• Backported from Alchemy (Rye Mutt's code) a PBR shaders change to "fix for divide by zero when haze density set to zero".
• More thoroughly rebuild draw pools and objects draw info when toggling PBR, to reduce the occurrences of "vanishing objects" (may sadly still sometimes happen, for other reasons, due to a still-to-diagnose-bug in LL's PBR code; a camera-zoom-far/wait-a-few-seconds/reset-camera-zoom workaround is still needed in these cases).
• Now automatically reset objects visibility when toggling "View" -> "Highlight transparent" to fix some transparent objects still not rendering.
• Implemented a workaround for the new SL server bug causing your avatar to always be flying at login. The toggle for this workaround is "Fix bogus flying state at login" in the "Advanced" -> "Character" menu.
• Added support for the "noon_pbr" environment setting preset in the ApplyDaySettings() Lua function.
• Added tool tips and highlighting support in the LSL script editor for the new llGetLinkSitFlags(), llSetLinkSitFlags() functions and SIT_FLAG_* constants.
• Linux only: updated libSDL to 2.30.2 and recompiled libndofdev against it.
• Linux only: recompiled Dullahan against the latest CEF 123 release.
• Minor code cleanup.

Enjoy !

Here are two new releases of the Cool VL Viewer, available now from http://sldev.free.fr/:

Cool VL Viewer v1.32.0.18 (stable branch):

New in this release:

• Added more NULL pointer checks to the octree code to try and fix a super-rare/random crash bug (that is sadly impossible to reproduce).
• Fixed a P2P voice signaling bug introduced in v1.32.0.16.
• Rewrote the voice device settings floater from scratch, with device display names used in combos instead of device Ids, added "Cancel" and "Apply" buttons (making things clear when something is applied or cancelled), and made it work properly (instead of sometimes reverting devices choices).
• Made plugins debug messages more useful (with plugin identification).
• Added a new SetVisualMute() Lua function to allow controlling the jelly-dollification of avatars via Lua. See the updated Lua manual for details.
• Code cleanup.

Cool VL Viewer v1.32.1.0 (new experimental branch):

This new viewer branch will be used to implement and test WebRTC voice and upcoming PBR renderer changes (mirrors, PBR terrain).

This week, the only change when compared to v1.32.0.18, is the preliminary implementation of WebRTC voice; sadly, the latter is for now not working properly, with spurious disconnections after 30s for all OSes, and with audio devices issues and WebRTC library deadlocks on disconnections under Linux (which I could make harmless to the viewer, by implementing a WebRTC plugin to communicate with that library instead of linking it to the viewer binary, plugin which is just relaunched in case of crash or freeze).

But at least this release will allow voice users to verify that it did not break Vivox voice (since some changes impact both types of clients): please, if you do use voice, do check that nothing got broken for Vivox when compared to v1.32.0.18, and report any breakage.

Enjoy !

Here are two new releases of the Cool VL Viewer, available now from http://sldev.free.fr/:

These are maintenance/bugfix releases and recommended updates.

Cool VL Viewer v1.32.0.19 (stable branch):

New in this release:

• Backported from LL's viewer-material-featurette a fix to rigged variant shaders environment-related uniforms not being updated properly after an environment settings change.
• Fixed a couple issues with MFA login (encrypted MFA hash not properly reset when the computer unique Id changed, MFA token not properly reset and its input box not hidden when you changed account selection on login screen after a failed MFA login on another account). The MFA token input box on the login screen is also now automatically focused after the MFA token refresh alert box is closed.
• Backported from LL's viewer-material-featurette the adoption of Blender's fork of MikkTSpace.
• Linux x86_64 only: updated jemalloc to the latest git version.
• Some code cleanup.

Cool VL Viewer v1.32.1.1 (experimental branch):

Same changes as in v1.32.0.19, plus:

• Fixed many issues in WebRTC voice code and backported LL's webrtc-voice viewer latest changes and fixes: it should now work (*), even though you might need to restart it manually just after login (I still need to fix this). You may test it in Aditi, and if you do, please do report any issue you would encounter with it. Added a new "DebugWebRTCVoice" setting to allow enabling automatically the "Voice" and "Plugin" debug tags when WebRTC voice is initialized on viewer startup.
• Backported from LL's viewer-material-featurette support for 2K textures (for now only enabled when you log in Aditi, and only usable in the 2K textures sim there).
• Added a new "TextureMaxResLevel" setting allowing to limit the maximum downloaded and rendered texture size. There are three levels: 0 = 512x512 pixels max, 1 (default) = 1024x1024 max, 2 = 2048x2048 max (grid permitting). Note that for GPUs with a low VRAM amount, limiting to 512x512 can be a life saviour while not really impacting the rendering quality (it is best to limit all textures to 512x512 than to have high discard bias values as a result of VRAM exhaustion, the bias level impacting all textures and not only the large ones).
• Backported from LL's viewer-material-featurette some preliminary code (no user-level change) for PBR terrain and mirror implementation.

Enjoy !

---

(*) Note that on some Linux systems, for WebRTC to discover and properly list all audio devices in the device settings floater, you might need to add defaults.namehint.extended on and perhaps also defaults.namehint.showall on to your ~/.asoundrc file (and create one with this/those line(s), if it does not exist); i.e., from a terminal, simply type: echo "defaults.namehint.extended on" >>~/.asoundrc; echo "defaults.namehint.showall on" >>~/.asoundrc. This (one time only) configuration step will be automatically performed in next release (via the cool_vl_viewer wrapper script), but I am interested to know if you do need to add the second line (defaults.namehint.showall on) for your Linux distro.

Here are two new releases of the Cool VL Viewer, available now from http://sldev.free.fr/:

These are recommended updates.

Cool VL Viewer v1.32.0.20 (stable branch):

New in this release:

• Hardened the reflection maps against potential crash bugs on objects destruction (now properly use reference-counted pointers on objects and spatial groups used by reflection probes).
• Fixed an issue with reflection probes octree intersection introduced by my fix to reflection probes selection code cleanup (oops !).
• Backported from LL's viewer material-featurette a fix to XMLRPC decoding of LLSD arrays, to support them properly in login parameters.
• Backported from LL's viewer material-featurette a fix to exposure maps being mangled with in-world exposure for PBR material previews.
• Backported from LL's viewer gltf_development a "fix to black Moon" in PBR rendering mode.
• Backported from LL's viewer a fix to objects LI costs not always being reevaluated after PBR parameters changes on their faces.
• Backported from LL's viewer a better way to pause reflection maps updates during environment settings transitions.
• More thorough reflection probes rebuild on parameter changes, or when manually triggering it via "Advanced" -> "Rendering" -> "Refresh reflection probes".
• Updated the boost libraries to v1.85.0.
• Patched colladadom to cope with the latest boost deprecations and recompiled it against boost v1.85.0.
• Linux x86_64 only: updated jemalloc to the latest git version, with a fix for a potential infinite loop on memory purging (never seen by me, but better safe than sorry).
• Some code cleanup.

Cool VL Viewer v1.32.1.2 (experimental branch):

Same changes as in v1.32.0.20, plus:

• Fixed the issue with bad connection to WebRTC voice on login.
• Adopted definitively my WebRTC plugin model and removed the compilation option to disable it.
• Backported from LL's viewer webrtc-voice more fixes to voice parameters changes detection on region/parcel changes.
• Backported from LL's viewer webrtc-voice support for automatic gain control, echo cancellation and noise suppression (see the new settings in the "Preferences" floater "Voice chat" tab, when logged in in a WebRTC-voice enabled sim).
• Implemented a way to flag special textures (e.g. future PBR terrain ones) as "mega textures" to allow rendering them at full resolution, even when all other textures have been limited on purpose to a lower resolution via the "TextureMaxResLevel" setting.
• Resynced the PBR shaders with LL's viewer material-featurette.
• Backported from LL's viewer material-featurette support for PBR high resolution probes, AKA mirrors. You can toggle them via the corresponding entry in the "Advanced" -> "Rendering" -> "Deferred rendering" menu. New debug settings of interest (make sure to toggle mirrors rendering off/on after changing them for full effect) are "RenderHeroProbeResLevel", "RenderHeroProbeDistance", "RenderHeroProbeUpdateRate" and my GPU over-usage work-around related "RenderHeroProbeCutoff". (*)
• Updated llwebrtc to match LL's latest version.
• Linux only: implemented in the cool_vl_viewer wrapper script the auto-configuration of ALSA for proper audio devices detection by WebRTC (i.e. the script now automatically updates ~/.asoundrc if needed).
• Some code cleanup.

Enjoy !

---

(*) Note that I find the current implementation rather disappointing, to say the least... First, anything but 2048x2048 mirrors (i.e. with "RenderHeroProbeResLevel" set to 3 in the Cool VL Viewer) will look super-blury. Second, mirrors cause a huge GPU power over-consumption (easily twice the consumption compared with mirrors disabled), with a massive impact on frame rates, even when mirrors are no more in the camera field of view (I partly worked around this with my "RenderHeroProbeCutoff" debug setting, but it is still not satisfactory; more work on it is apparently ongoing on LL's side). Finally, mirrors are hard to setup (weird constraints with the mirror reflection probe, its centering on the mirror plane, and its size and rotation relatively to the latter).

Here are two new releases of the Cool VL Viewer, available now from http://sldev.free.fr/:

These are recommended updates.

Please note that if you did not try it already, it is about time to give the v1.32.1 experimental branch a go and report any issue you would find with it, before it gets pushed to the stable release status (which might happen soon, depending on how fast LL will push their "featurette" and/or "webrtc" viewers to release).


Cool VL Viewer v1.32.0.21 (stable branch):

New in this release:

• Do not reset the TOS and critical message acceptation flags after a failed login (e.g. because of an expired MFA hash), to avoid getting stuck in a "try login/get TOS or critical message/accept it/fail login/rinse and repeat" process.
• Fixed an off-by-one buffer overflow that could have happened with invalid primitives-related UDP messages. See this thread for details.
• Fixed the "black Classic clouds" issue seen in PBR rendering mode with the new (hacked) PBR environment settings (e.g. PBR midday), due to the fact those settings got an ambient light set to black (!). Added a new "CloudsMinAmbiantThreshold" debug setting governing the ambient light intensity level below which the clouds are tinted following the Sun or Moon light instead of the total ambient light.
• Fixed a bogus terrain texture offset calculation in LLDrawPoolTerrain (it was using the agent's region position instead of the position of the region to which the terrain belongs).
• Properly delay asset storage fetches when they get fired up before the agent region is set on login (race condition fix).
• Backported from LL's featurette viewer a new algorithm to reserve texture indices in PBR rendering mode, and improved it to automatically disable reflection probes when not enough of them are available.
• Backported from LL's featurette viewer a fix to the GLTF importer/loader dealing with the ORM map.
• Backported from LL's featurette viewer a fix to a failure to erase the GLTF overrides on cache entry removal of objects using PBR materials.
• Changed "RenderSkyAutoAdjustBlueDensityScale" default to 0.8 instead of 1.0, for a better looking sky in PBR rendering mode with auto-adjusted legacy EE settings.
• Moved the broken "Screen space reflections" PBR renderer setting from the Graphics preferences tab to the "Advanced" -> "Rendering" -> "Deferred rendering" menu, and added an explanatory alert box when enabling it. Since LL does not seem in any hurry to fix it, just consider it is a dead feature...
• Added several entries to the Advanced" -> "Rendering" -> "Type" menu.
• Build prerequisites raised to Python v3.3 minimum (instead of v2.6).
• Removal of dead code and cleanup.

Cool VL Viewer v1.32.1.3 (experimental branch):

Same changes as in v1.32.0.21, plus:

• Backported from LL's featurette viewer PBR terrain support, with the exception of the experimental feature for generating terrain normals. (*)
• Added support for rendering the PBR terrains base color texture when not in PBR rendering mode (i.e. you will not see grey terrain in ALM or forward rendering mode, unless the base color texture itself is not used by a given PBR terrain).
• Backported from LL's featurette viewer the latest changes and fixes to mirrors support with, in particular, the new distinction between static mirrors (only non-moving objects reflected) and dynamic mirrors (all objects reflected). Note that all old mirrors will be considered static, so you might want to change their reflection probe for "Mirror (dynamic)" in the reflection probe edit floater.
• Exposed the "TextureMaxResLevel" debug setting values as three new entries in the "Advanced" -> "Rendering" -> "Textures" menu ("Fetch/render at NxN pixels max" with N = 512, 1024 or 2048).
• Some code cleanup.

---

(*) The corresponding commit mentions "Proof-of-concept PBR terrain normal gen feature flag. Final implementation would likely be very different.". So no need to implement something that will likely change a lot or get removed...


Enjoy !

Here is a new release of the Cool VL Viewer, available now from http://sldev.free.fr/:

This is an emergency bugfix release. All users of v1.32.0.21 or v1.32.1.3 (EDIT: and v1.32.1.4) shall update.

Cool VL Viewer v1.32.0.22 (stable branch):
Cool VL Viewer v1.32.1.5 (experimental branch):

New in these releases:

• Reverted the backport of the "new algorithm to reserve texture indices in PBR rendering mode": this caused disappearing alpha-textured objects in PBR rendering mode with shadows on. See this thread for details.
• Reverted the backport of the "fix to a failure to erase the GLTF overrides on cache entry removal of objects using PBR materials": when such an object re-rezzes in the same session (e.g. due to draw-distance and camera FOV changes), its PBR overrides were lost (the sim do not resend them as long as the viewer does not reconnect to the region).
• Fixed a copy/paste bug which caused the new "Advanced" -> "Rendering" -> "Type" -> "Full bright" option to actually toggle the "Alpha mask (full bright)" one.
• Fixed a typo in the new "CloudsMinAmbientThreshold" setting (miss-typed as "CloudsMinAmbiantThreshold").

Enjoy !

---

Note: I did not build aarch64 releases, since the major bugs that have been fixed only impact PBR rendering and ARM64 SBCs cannot currently do it...