Beta Release 56

#1

Beta Release 56 includes up to Developer Release 7208. You will need to update to this version in order to visit domains running it. Should be available to you within about an hour. Here’s what’s up:

Notable Changes

Technical Issue Fixes

  • Make rotation using Vive controller work
  • NearParentGrab module now sets the correct previous parent id when switching grabbed overlays from one hand to another
  • Fix race condition accessing QString sub-properties in zones
  • Use selection model of TreeView
  • Fix polyline crash and flicker
  • Rework scriptengine lifetime
  • Avatar Attachments now render correctly
  • Fix the web surface of the tablet
  • Ensure thread-safety for model pointer access
  • Fix sorting on import menu
  • Fix grab entity methods + jshint cleanups
  • Fix typos in controller modules disabler
  • Add resizable flag. Use the flag for Create only
  • Fixed Google Blocks download issue
  • Change tablet spawn position
  • Shapes now render at the correct size.
  • Fix OpenSSL build instructions (related docs)
  • Flag node as local by IP from UDP header, not IP from packet data
  • Fix Record app’s toolbar button not highlighting when app dialog open
  • Optionalize the debug spam of makeUserConnection.js
  • Fixing/improving and cleanup tablet inputs
  • Audio devices split (related docs)
  • Pull back the stylus position in hand.
  • Don’t let scripts control visibility of laser pointer overlays
  • Don’t add items to the scene graph until they have a valid parent chain
  • Fix up the doppleganger mirror script
  • Add Input Field On Top of Keyboard
  • Remove old versions of the VC Runtime from the installation directory
  • Fix RunningScripts in tablet mode
  • Fix dimensions of rendered web entities
  • Fix for TextEntity rendering; the translation was incorrect
  • Fix alpha on Line3DOverlays with glow
  • Teleport.js: can now teleport small distances with small avatar scale
  • LODManager uses batchTime and engineRunTime for automatic LOD adjustment
  • Fix EntitySelectioTool and improve far rotate
  • Create Mode: the keyboard tray is not styled correctly
  • Don’t constantly reload animGraph if URL didn’t change
  • set grabbable and other checkboxes correctly if userData is blank
  • Fix mirror and fade threading crashes
  • Fixing the broken “on hud” overlay’s render transform
  • Fix ModelOverlay visible change
  • Fix pointer event properties
  • Laser Pointer fixes and updates
  • Modal dialogs async fix
  • Move Tablet becomes toolbar menu from General to Developer menu
  • Correctly grab tablet in 3rd person
  • Made handControllerPointer into modules in hope of preventing mouse cursor interference with tablet
  • Remove asserts that break Debug build
  • Text3D Overlays and Entities fixes and no-AA
  • Fix parent grab
  • Fix dde visibility
  • Fix long tablet lag on first load
  • Unexpected Behavior when pressing Enter on Input Field
  • Fix auto dropping and re-grabbing loop
  • Fix near trigger
  • Moving transform evaluation out from render loop to gameplay loop / fixing tablet tearing

Known Issues

This release includes significant changes to several major systems, including controller grab, rendering, and the control tablet. As of this release several issues remain:

  • Issues transferring from far grab to equip
  • Scrolling the tablet with lasers
  • Equipping with the grip

New Documentation

Updated Docs for Beta Release 56

Build Guide: OpenSSL build instructions fixed.
Entity Properties: Adding ability to scale unlocked entities that can be grabbed or dynamic.

Other Doc Updates

#2

Still highlighting
Still highlighting.jpg1920x1005 930 KB

picture says all the words

#3

Awful
Awful.jpg1920x1005 993 KB

let’s make this clear when you’re moving your camera around how do you think this is going to work out for people when they can’t disable this

and how long is this going to take you to fix this

#4

My avatars textures stopped to render after this update… (They are not embedded in the FBX.)
I used that for months. Now it’s broken.
(And don’t ask me to embed them. If I edit it in Blender I get the rigging devastated.)

Non-embed textures work fine for fbx Models… it seems to affect only the avatars.

image
image.jpg703x732 147 KB

Here a good example… My avatar that is not rendered as before… and the ground which is a non-embed texture FBX work fine.

#5

I noticed loss of textures on avatars with this release too… but it was odd. same person, same avatar, teleporting away and back and then logging out and back in. I was not able to make a connection between cause and effect.

#6

I have looked into this as this also hurt a few of my own avatars. At this time, the client is ignoring textdir. To resolve the issue, move the textures to the same location as the fbx files. You can further find out where the client is trying to find the files by checking the logs.

#7

My assignment-client stopped building on my Debian Server:

/home/menithal/Code/hifi/assignment-client/src/assets/AssetServer.cpp: In member function ‘virtual void AssetServer::aboutToFinish()’:
/home/menithal/Code/hifi/assignment-client/src/assets/AssetServer.cpp:283:49: error: ‘class QThreadPool’ has no member named ‘tryTake’; did you mean ‘tryStart’?
         auto pendingRunnable =  _bakingTaskPool.tryTake(it->get());

Also Scripts keep removing them selves when I refresh them.

This is very, very annoying as it makes working on client-scripts a pain

#8

https://blog.highfidelity.com/automatic-creation-of-compressed-assets-by-the-high-fidelity-server-e887b7da4c11
Am I right in thinking the asset baker only bakes atp stored assets?
is this just converting to ktx or is it doing the Draco conversion also?
There was a baking tool floating around on github that will bake out an entire domain, is this still underdevelopment?
I would like to bake all of Cromptonmoor
helllp lol

#9

You are right that it only bakes ATP assets.
It’s also doing Draco.
No idea on stand alone baker - it’s still in code base, but hasn’t been mentioned in a while.

Baking takes just short of forever. If you remember the old standard apartment scene… you know the default content from ages back with the robot cleanup script and all that… it took around 5.5 hours for it to bake. So… if your content is in ATP it will, eventually (*maybe) bake.

*maybe caveats.
If FBX comes from Maya/MayaLT it will likely fail to bake as Maya embeds some DDS format textures that define PBR light environment and baker chokes on those. That’s due to be fixed.

If you get a baked version you may notice it no longer looks correct – this is usually due to Draco having mangled the mesh and screwing up how the UV map “fits”. This occurs mainly on small models with need for very precise alignment of mesh to textures. Solution = go to ATP browser and click the bit that toggles used baked version and it’ll revert to not baked/not compressed (Draco) version.

If I understand the entire thing correctly – it all eventually becomes KTX for textures whether you pre-bake or not. You’re trading your server time doing the bake vs having client download a not KTX and convert on the fly, which - shoving off to the server to do over time vs hammering client in real-time, really is the right thing to do (in most cases).

#10

Some time back they made a change that only looks for FBX textures in same directory as FBX was located – something I lobbied heavily against as it would break avatars when it first came up a few months back. Unfortunately I wasn’t around due to hurricane when it came again and was implemented. Solution - your AV textures must be in same dir as FBX vs the old way of FST pointing to an AV FBX + a relative tex dir.

#11

The standalone thing which i have somwhere

would take the exported json file thing with all the models
download and convert everyhing and ammend the json thingy to the new urls once uploaded
it was kinda tricky but clever and how i made the drax thing i made all ktx in one go
like u said it gets the fans spinning on the pc but its worth it for automated conveniance
i hope they keep it as a stand alone thing
i cant find where i downloaded it from so it may be outta date

found it
https://github.com/highfidelity/hifi/pull/10915

#12

Yeah - they gave links to it a couple of times when it was in development (In the PR system), but, AFAIK, only way to get a more recent build is to self-compile.

#13

I saw that error thrown when I tried to compile against Qt5.7.x (I had Debian 9 and thought it had 5.9 but discovered it was 5.7 after seeing that – then moved to dangerous lands of bleeding edge Deb 10 land where 5.9 is default Qt) vs the now required 5.9.x – that’s only thing I can think of to explain as I’m getting a clean compile on Ubuntu 16.04, Ubuntu 17.10 and Debian 10.

Also - I just remembered… if you passed the path to Qt via cmake with a -DQT_CMAKE_PREFIX_PATH that is being ignored in favor of environment variable QT_CMAKE_PREFIX_PATH… that messed with me on one of my build boxes as I had env set for Qt5.6 and -D option at 5.9 – I saw a PR to fix that, I think, but not sure if that got rolled into this RC. To be safe, make sure you set env var to proper Qt location and pass it as -D option to cmake… that seems to be the current 100% solution.

Of course then you can run into problems with AC not running due to NVTT issues (depending upon your “luck”). On Ubuntu 17.10 and Debian 10 it claims NVTT can’t be found even if you make sure it can be found… drilling down you’ll see one NVTT lib is found, but it then has no reference to another NVTT lib it needs. Solutions = $LD_LIBRARY_PATH hackery or moving NVTT libs to something like /usr/local/lib then sudo ldconfig (insuring /usr/local/lib in ld paths) or – my choice – patchelf and chrpath hackery to add rpaths such that I can place all this stuff in compartmentalized locations and not “cross pollinate” base system with hf specific libs.

#14

having a converter that uses the atp is stupid
because nobody uses it its a stupid as having a apple build but not a linux one
who uses apple for vr?

#15

I recently discovered a bug that let me view the atp of any domain I travelled to. Lots of people use atp. Myself included.

#16

I saw that you were in contact via PM with staff about it. Thanks for reporting it!

#17

A hotfix for the issue of missing textures–the High Fidelity client ignoring texdir–will be live in about an hour. Thanks, everyone, for reporting it so quickly and including screenshots and details.

#18

@Judas. I agree and also don’t. The problem with Linux is of which distro(s) to support. Traditionally it isn’t up-stream’s job do do distro-specific packaging (though ideally upstream should be working with each distro’s packager to make sure things are workable).

Now, however, we have FlatPak, which makes things a little different. I’d love to see an official compiled and tested FlatPak for any FlatPak-supporting distro. Especially useful at this beta stage of development when finding a suite of competent packagers out in multi-distro-land may not be feasible.

Contrary to the popular belief, I - and many Linux users - DON’T like compiling code! We use Linux because it is easier (for us) than the commercial offerings, not because we like to make life harder for ourselves! :grinning:

Also. How is the Android Interface going, anyone? I rather like the idea of dropping Interface on one of those cheap Android STBs from Asia as a dedicated VR terminal without any hassles at all! (Not that Android is the ideal platform for that, but it is nice and generic and very available.)

#19

Since the build 56, The new fbx I created with embedded texture are rendered in fluo in HF.
Old models are still correctly rendered, and the fbx with relative texture (non embedded) look great (see on the right)

image
image.jpg1249x710 492 KB

In both cases, it’s store on the web (Not in asset server)

Anyone have an idea what setup I might have changed by mistake?

#20

That’d be https://github.com/highfidelity/hifi/pull/11536 (PR11536). I spotted it about a month ago when optimizations to CMAKE were introduced, and thought it’d have been fixed. It was pulled into the master as of 4 days ago, so -DQT_CMAKE_PREFIX_PATH should work again.

As for the NVTT situation… I’ll have to take a look. Could something like -DNVTT_PATH help out at all (doesn’t exist for now but just seeing if that’d helpl)? I haven’t had issues compiling on a Ubuntu 14.04 system, but I’ve been having issues with the stable version being somehow out of date. I’ll have to compile from master and see what is going on (and when I should update to 16.04).