• We should add a visual test with the mesh we use in the parcel deployed in ZONE 58,-71, it has different materials with different propeties representing standars configurations

https://play.decentraland.zone/?position=58%2C-71

Also we should add:

• Visual test for PBRMaterial: A transparent plane with alpha == 1 with opaque objects behind and SSAO enabled. SSAO should not be visible.
• Visual test for PBRMaterial. At least emission and alpha properties with different values (AlphaTexture set).
• Visual test for PBRMaterial. AlphaTexture and AlbedoTexture both set with the same texture. The texture has to have different RGB values (i.e. be colored with red green and blue) (to prevent regressions from #283)
• Visual test for GLTF with transparencies that come from albedo (0.5) and NO alpha texture (to prevent regressions from decentraland/explorer#2328)
• Visual test for GLTF with transparencies and no emission.

• To create an XML that is parseable by CircleCI to display the checklist in human readable format

This two buttons should be visible on the taskbar but giving visual feedback that both are locked. After clicking them, a small toast reminds the players that they should create an account or have an authenticate name. Each tip contains the link to unlock them.

We have 2 classes for Material to do practically the same things. We need to extract the common code into a common library.

• Create Material component visual tests before implementing this refactor

Sometimes when you do hot reload, one or several entities get red tint, as if they were outside the scene bounds. The entities are tinted red but they don't have their bounding box marked in red.

This issue is not easy to reproduce, happens inconsistently, likely a race condition. It always gets fixed by manually reloading.

It would seem that in some cases an entity gets first loaded in 0,0,0 and then is still considered out of bounds when positioned in its right place.

We have the theory that this same problem happens occasionally in production, where some entity is randomly not visible in a scene. There were various random irreproducible instances reported in Halloween, where different entities just didn't appear, but reloading fixed it. If this is happening in production, instead of getting the red tint, the entity won't render at all.

How to reproduce?
Take any large scene, and make minor changes so that it hot reloads. After a few (maybe as much as 10 attempts), you'll surely have some entity appear red, as if out of bounds.
For example:
https://github.com/decentraland-scenes/Genesis-Plaza

When another player's avatar position changes, the walk or fly animation is executed. This happens even when the avatar is being moved by a moving platform. This also causes any other animation (such as dancing) to stop. Ideally, we would want other avatars to behave as the own avatar, when being moved by a platform.

Due to some optimizations we are using the same AvatarEditorHUDController in different tests. The benefits are obvious, a single initialization of a time consuming UI creation, but due to the current SetUp/TearDown approach, the instance of the controller is not getting collected by the garbage collector and not being disposed.

This "ghost" instance is subscribed to Catalog and reacting to changes on it. Although right now it's not happening, it could lead to some issues in the future whether somebody want to add mock wearables to the Catalog and making the controller instance crash by reacting to them.

To solve it we can use SetUpFixture https://nunit.org/docs/2.6/setupFixture.html. We can instantiate the controller in the Fixture SetUp, run all the tests in the same namespace using the shared controller, and dispose it in the Fixture TearDown.

AvatarEditorHUDControllerShould and AvatarEditorHUDViewShould tests are subject to this improvement, but a project-wide research is needed to check that no instance is ghosting in the background.

I found a small issue with movePlayerTo , after using it the performance drops (using it more than one time don't make it worse, the effect does not accumulate)

Tested on Chrome, GNU/Linux, using the latests ECS version 6.6.0

Before:

After:

Code:

const respawner = new Entity()
respawner.addComponent(new BoxShape())
respawner.addComponent(new Transform({ position: new Vector3(8, 0, 8) }))
respawner.addComponent(
  new OnPointerDown(
    (e) => {
      movePlayerTo({ x: 1, y: 0, z: 1 }, { x: 8, y: 1, z: 8 })
    },
    { hoverText: "Move player" }
  )
)

engine.addEntity(respawner)

Test scene, in case that portion of the code is not enough to reproduce the issue:
slow_after_teleport.zip

Currently the base code makes testing this feature a mocking nightmare. It will be done after decentraland/explorer#1322 is completed.

When I add a Material to a PlaneShape with a static image Texture, the UVs are flipped along the vertical V axis compared to adding the same material with a VideoTexture.

Example here (click on right screen to start video):
https://bug-report.vercel.app/

code:
https://app.zenhub.com/files/220994476/2ae72b64-6c22-4ec4-b3ec-47c0757e8083/download

We need to do an assessment of all the tests that make use of the old IntegrationsTestSuite_Legacy suite to the new IntegrationTestSuite

The result of this tasks should be a list of all the classes / services / components that needs a refactor and an estimation to be achieved.

The BoxShape component can't have its uvs set, to determine how a texture gets wrapped around it.

It would be great to be able to set it just like we do for the PlaneShape, where you can use that to stretch, rotate or tile a texture how you like.

Would be great to also be able to map the inside of the cube, to use as a skymap

Would be great to also have them in other shapes like SphereShape

Related comment from someone in discord:
Please could we gain the option to reverse the texture normals on a primitive SphereShape as the usual -1 scaling tricks do not workaround.
Spheres mapped inside out and viewed from the inside outwards would allow scenes to lever skybox like qualities along with certain essential perks like external/dynamic video mapping. Something which afaik gltf's alone cannot currently match.
The platform already has everything else in place in this regard... only requires reverse mapping to complete.

Due to the lack of context of explorer.json; the field "renderProfile" doesn't communicate well what it does.

• It's a 0. Could this mean a boolean value? Is this a disabled feature?
• Does the "profile" there mean that this disables rendering of other user's Profiles?
• After taking a look at what it does, it sets night mode as a default when it's set to 1.

I think this is going to be super tricky to understand in the future. After taking a look at what it does, it's an "aha! it's the default rendering profile to be used!". Let's fix this for more understandable configuration:

• Change the name from "renderProfile" to "defaultDayNightMode", or "defaultEnvironmentalRenderingSetup" (avoiding Profile to prevent confusion with avatars or other people, avoiding Settings because of all the settings the user can select) for the name of the configuration, if we're looking to have more rendering profiles in the future.
• Change the value from 0 or 1 to "day" and "night" (make it an enumeration in the config, not in the code)

TL;DR: Pay tech debt, turn this

{
  "renderProfile": 1
}

Into this:

{
  "defaultDayNightMode": "night"
}

So it's more understandable for someone modifying the client

right now the "event-triggered" tests are explicit (disabled) as they fail even though the events work.

The comment on the tests mention that the PointerEventsController has trouble getting the camera in its RetrieveCamera() but i just debugged a little and I think that's no longer the case but further investigation is needed.

• Fix "event-triggered" tests and remove the explicit tag and category

With voice chat enabled, the mic icon does not show next to the speaking user and it's very hard to detect where the audio comes from

Currently the AvatarShape/Renderer/Wearables tests are loading a catalog.json from TestResources.

This catalog is pointing to the assets in production. If something changes in production (like a hash) all the tests will break. To prevent this, a basic set of wearables/bodyshapes must be downloaded and included in the repository and the catalog must point to them.

Some images used as textures show up as red question marks on white, rather than the correct image.
It appears to depend on the wording of the folder name/path containing the image, or on the file name.

I can reproduce it if the path contains a segment of "advertise" or if the filename starts with "YourAdHere".
This does not require my ad blocker to be on for the site, or even enabled, and happens both in preview and in-world.

https://gitlab.com/dcl-tech/texture-issue exemplifies the issue, and the game.ts has commented out lines that show what does and doesn't exhibit this for various filenames of the same file.
I have also seen it where the filename would work, but a segment in the path is "advertise"

Seems to be not too hard to work around once you understand how

• Reactivate Explicit test for AvatarRenderer

Right now, we lack a proper AvatarRenderer testing suite. We should re-organize the class and mock anything necessary to get to a deeper coverage. The aim of this task is to make the avatar loading logic more resilient and prevent invisible wearables bugs and such in the future.

In kernel's side we should add a new opions type, like the OnPointerUUIDEventOptions, and a new constructor overload in the 3 OnPointerUUIDEvent classes (detecting inside which options where used somehow)

more info in the one pager:
https://docs.google.com/document/d/1I2bcntDm3KIp2rO_J7cNIYQ94D2zrEG2-jY92wpfgBk/edit#heading=h.lsdbyfkkfvem

Brian:

This can happen in:
ORG -55, 142
ORG 12, 77

As far as I debugged, the builder scenes with smart items have some await statements that makes the scene send incomplete initialization messages. Waiting a bit before sending the remaining messages with all the addEntity and addComponent necessary to complete the scene.

If I force a setTimeout(1000) after the first eval statement in worker code, the remaining messages to load scenes are magically sent. In this interval, no messages come from systems or events, so it should be an await.

Before the rest of messages come, suspicious chest connect logs can be seen.

All the non-smart item objects should be added to the scene in the first frame or the loading lifecycle will fail.

@cazala @moliva

Seems to be that when the chat history gets quite long, every new chat message received takes some milliseconds in processing, and that causes noticeable hiccups

Recently, Meteor Chaser suffered from a change that made 1/3 of the meteors stop appearing. After debugging with Rob, I think we found the root.

It's a common trick to scale an entity down to 0.01 scale to keep it hidden before you want to show it. The meteors have a model and animations that take a second to load, so what Rob was doing was he first kept the meteor loaded but super small while it loaded, then enlarged it and animated it.

As a suboptimal workaround, he started to render the meteor up in the air at full scale, but it's quite visible at night.

My suggestion is to reevaluate culling of entities every time there's a scale change.

Some excepts from my conversation with Rob:

Carl reports that he found some unexpected invisible colliders in one of his scenes.

He made this video to show the bug:
https://youtu.be/tvcbvSTtseg

This should help with the dizzyness issue.
https://www.reddit.com/r/Guildwars2/comments/10gztt/a_solution_for_those_that_experience_dizziness_or/

Users reported "X" as being very helpful. We should mention that when the user tries to skip the tutorial

Maybe a second dialog that lets the user know of this dialog...

Task list:

• Review existing scene metrics criteria

  • Write RFC and share with the team

• Gather existing scene metrics

  • Meet with data to obtain quantitative metrics usage
  • Remove material replacement from bounds checker red feedback
  • Add support for texture counting to scene metrics counter
  • Write tool to perform qualitative analysis
  • Implement new memory metrics for textures, meshes, audio clips and animations (WIP)
  • Add support for Billboard counting to metrics
  • Add support for TextShape counting to metrics
  • Fix NFT shape tests, add tests for gif NFTs texture counting.

Easy to test using TEST_WEARABLES / ALL_WEARABLES flags.

The glowing materials are too saturated in the world.
After talking about this issue to @ShibuyaMartin it seems like the world renderer is multiplying the glowing factor much more than expected.

In order to be consistent with what our designers create, we should have a kind of glow adjust model showing many different combinations of colors, textures, and emissivity. Then we should compare how this adjustment model is rendered in different model editors and inside the world, and we should adjust our default values to match as close as possible the editors. We must consider at least https://sketchfab.com/, and possible https://www.blender.org/ and https://www.babylonjs.com/.

This only seems to happen in preview and with relatively large scenes for some reason, or scenes that have something particular in them, haven't found the pattern

If I run this scene once, it sometimes loads fine. If I point at things I see feedback.

Change something in the source code for hot reload to kick in, and the feedback is gone. Same item, same distance, same everything.
If I click the entity, it will execute its click behavior correctly, I just don't have the hover feedback. If I reload the browser manually, it often gets fixed.

With this scene, I see it happen 100% of the time with all clickable entities.

https://github.com/decentraland-scenes/Xmas_Plaza

Check the standalone branch of this repo for the most updated version.

(ask me (Nico E) for access if you can't open)

I'm using a Mac & Chrome when I see this happen.

Be able to create an NFT for both male and female, e.g. with very similar design and slight gender differences, and that display two different gender versions for the thumbnail.

   thumbnail : "dcl:://asdasdasd/generalThumbnail"
   representations : [
     {
        id : "BaseMale"
        thumbnailOverride : "dcl:://asdasdasd/maleThumbnail"
     }.
     {
        id : "BaseFemale"
        thumbnailOverride : "dcl:://asdasdasd/FemaleThumbnail"
     },
     {
        id : "Robot"
        thumbnailOverride : ""
     }
  ]
}```

![image.png](https://images.zenhubusercontent.com/5d91dca0703cca0001efbe30/99c5f190-de1b-4285-bc73-09910249ce8b)

If I play an animation and then I want to play another animation, or the same animation again, I need to first explicitly stop the last played animation. It's a very common newbie pitfall, they aren't able to play an animation a second time, and the reason is that they need to stop the animation explicitly before they call it again.

This doesn't work

sit.play()

Input.instance.subscribe('BUTTON_DOWN', ActionButton.POINTER, true, (e) => {
  sit.play()
}

When dealing with a model that has several animations that may have ben called before, the workarounds needed for this can get really nasty.
See this example:
https://github.com/decentraland-scenes/Block-dog

Here you need to explicitly stop all these animations every time before you want to play another one
sit.playing = false
stand.playing = false
walk.playing = false
drink.playing = false
idle.playing = false
stand.playing = false

This issue has been around ever since we migrated away from Babylon. Back when HardlyDifficult did the escape room tutorial (April 2019), he avoided mentioning the workaround for this bc it was a bug we were supposedly going to fix by the time the videos were released. ....sure.

Right now, some scene music/broadcasting is not being heard from the contiguous scene. It would be desirable that we change that behavior (using 3D positioning of the sound). We could make it so that the scene.json can disable the music from neighbouring scenes (in case I have a noisy neighbor).

Opening external links is a restricted action that intentionally can only be triggered from a user's action to avoid spamming.

We allowed it to be called from OnPointerDown events and it seems also from click events
BUT, it's reported that it can't be called from a global E or F event.

This seems arbitrarily inconsistent. It's for example a good use case to be able to call an external link from an NPC interaction, and there it would work with E and F global events too.

As reported, this works:

Input.instance.subscribe('BUTTON_DOWN', ActionButton.POINTER, true, (e) => {
  openExternalURL("www.google.com")
}

This doesn't:

Input.instance.subscribe('BUTTON_DOWN', ActionButton.PRIMARY, true, (e) => {
  openExternalURL("www.google.com")
}

In areas with heavy scenes this issue can lead the user to think that the explorer got stuck, for example in
https://play.decentraland.zone/?ENV=zone&position=64%2C-65

Maybe we shouldn't show the "Authentication successful" or "communications established" message and just show the loading %s.

Whenever a new user enters the world for the first time, his avatar face snapshot is never loading and the avatar HUD stay in "loading" state indefinitely:

The only way to avoid this issue is opening the AvatarEditorHUD and click on DONE button. From that moment on the avatar face snapshot loads always correctly.

STEPS TO REPRODUCE IT:

1. Create a new Metamask account.
2. Enter the world with this new user.
3. Notice the issue.

NOTE: Debugging the code I noticed that the problem is that the profile that is being sent to Unity during the LoadProfile() event contains invalid snapshot urls. For example, something like this: https://peer-eu1.decentraland.org/content/contents/QmVrwSKEt4W4RovVhQbNTZxu4XNHfjVrs5QF3vjML59SaF. Then Unity try to resolve this url and fail.

This bug is being reproduced again. No clear steps or repro rate.

Specs: Windows + Chrome + GeForce gtx 1080 ti

This issue should cover the Making the click detection more “loose” section of the TDD:
https://docs.google.com/document/d/1I2bcntDm3KIp2rO_J7cNIYQ94D2zrEG2-jY92wpfgBk/edit#heading=h.c08r7aii51yq

Investigate the real issue behind: decentraland/explorer#1440 and decentraland/explorer#1441

The error is related to the order in which OnDestroy and Dispose are called.

Remember that Dispose it might not be called when you expect it when running tests (therefor the need of the CleanUp method)

One of our Users is reporting errors while trying to use a Marc o Matic wearable.

This is the log he's receiving in the js console:

An abnormal situation has occurred: the PlayerLoop internal function has been called recursively. Please contact Customer Support with a sample project so that we can reproduce the problem and troubleshoot it. 
(Filename:  Line: 377)
rollbar.min.js:2 OpenGL Error: Invalid texture unit! 
(Filename:  Line: 72)
this right?
inside the abnormal playerloop it says
r.<computed>    @    rollbar.min.js:2
_JS_Log_Dump    @    e62eb3f0-7954-483b-8567-d1ec963b89ef:8
(anonymous)    @    06dc05fe:0x20218b
(anonymous)    @    06dc05fe:0x2020ba
(anonymous)    @    06dc05fe:0x201da4
(anonymous)    @    06dc05fe:0x20123f
(anonymous)    @    06dc05fe:0x468271
(anonymous)    @    06dc05fe:0x46643c
(anonymous)    @    06dc05fe:0x478bf1
dynCall_vi    @    06dc05fe:0x1796fb3
(anonymous)    @    e62eb3f0-7954-483b-8567-d1ec963b89ef:8
dynCall    @    e62eb3f0-7954-483b-8567-d1ec963b89ef:8
dynCall_wrapper    @    e62eb3f0-7954-483b-8567-d1ec963b89ef:8
wrapper    @    e62eb3f0-7954-483b-8567-d1ec963b89ef:8
(anonymous)    @    e62eb3f0-7954-483b-8567-d1ec963b89ef:8

To add chinese support to DCL chat widget we need to change the default font to include the chinese glyphs.

Here is a link with the unicode ranges:
https://stackoverflow.com/questions/1366068/whats-the-complete-range-for-chinese-characters-in-unicode

https://en.wikipedia.org/wiki/List_of_CJK_fonts

Sometimes we ahve a Null reference when accepting a friend.

Because we are using WASD only -- people can't move around if they have an alternative keyboard

This will also fix:
decentraland/issues#299

When clicking sell CTA on any collectible item, it took the players to their assets URL on the marketplace when it should be taking them directly to the wearable URL instead.

We should identify the static parcels in kernel to terminate their workers after the code eval.

We can base the implementation on decentraland/explorer#1292

some profiling regarding this has been documented at https://docs.google.com/document/d/1alngddAcBxyOHQQ3ieuEGv2H7Y1GXR1LLxje05406sY

This component has 0 test coverage.

Since testing an AudioStream without mocking will be impossible, a refactor will be needed to adequate the class for mocking. A good approach would be using the Humble Object Pattern https://blogs.unity3d.com/2014/06/03/unit-testing-part-2-unit-testing-monobehaviours/

• Coordinates should still visible near the map cursor when the scene card is opened.

• For scenes without name or image, show the default image and "Untitle Scence" if those fields are empty.

• Remove coords and scene description from the card (let's discuss coords. May be we should keep this info if the card overlaps the cursor plus coords reference on the map).

• Add ongoing event tag (ray icon+"now" copy with blue background).

• If a creator of a scene doesn't exist, don't show it.