krpano 1.22 - Reactive APIs, ThreeJS 3D Plugin, krpano Maps, Improved Javascript APIs, Custom Encryptions

kme

I'm not asking to fix, I'm just reporting there is a difference in behaviour in 1.22 vs 1.21 and it is up to you to decide how much "backwards compatibility" you want to support. If the first was wrong, then maybe it should be adjusted in 1.21

More reports coming, but need to break them down to simplyfied code...

kme

Different handling of origin and scale:

http://krpano.kri-soft.be/playground/hittest/index.html

http://krpano.kri-soft.be/playground/hittest/index_122.html

Code

    <action name="enable3Dcursor" type="js" autorun="onstart" if="device.desktop"><![CDATA[
    
    var renderloop_id = null;
    var hit ;
    
    krpano.global.renderloop_id = krpano.actions.renderloop( renderloop );                

    function renderloop()
    {
    
            hit = krpano.actions.screentodepth(krpano.mouse.x, krpano.mouse.y);
            if (hit)
                krpano.layer.getItem("hitdata").html = "we are htting the cube";
            else
                krpano.layer.getItem("hitdata").html = "no hit";
            console.log(hit);
            
        }
    
]]></action>

Alles anzeigen

- the VIEW is totally different (scale factor?)

- the .obj is not moved to the same location as in 1.21 (scale factor not taken into account?)

- hittest is not accurate in 1.22: it is not on where the object is, but it does cause a hittest on where the object would be if it was in the same location as in 1.21 (move mouse somewhere below the cube and it will hit)

Tuur

kme links don't work

Tuur

spacerywirtualne

Zitat von klaus.krpano

Do you have an example?
The static layer-sublayers declarations and the dynamic parent changes are two very different things.

yes

not working

Code

<style name="poiset_main_scroll" type="scrollarea" parent="poiset_main_scroll_mask" direction="h" height="100%" handcursor="false" y="-20" x="0" zorder="1" align="top" onloaded="setcenter(0,0)" keep="true"/>

working

Code

<style name="poiset_main_scroll" type="scrollarea" parent="poiset_main_scroll_mask" direction="h" height="100%" handcursor="false" oy="-20" x="0" zorder="1" align="top" onloaded="setcenter(0,0)" keep="true"/>

my solution: y -> oy

It looks like the ‘setcenter’ in the scrollarea works differently in a 1.22 version

Piotr

kme

Zitat von Tuur

kme links don't work

they work, but only on desktop (if="device.desktop")

Tuur

Zitat von kme

they work, but only on desktop (if="device.desktop")

not for me on desktop

Tuur

klaus.krpano

Zitat von kme

I'm not asking to fix, I'm just reporting there is a difference in behaviour in 1.22 vs 1.21 and it is up to you to decide how much "backwards compatibility" you want to support.

I want as much backwards compatibility as possible of course .

I have changed that now - when a layer is defined inside another layer, then its parent setting will be ignored - as previous versions were 'ignoring' it too, but there more by accident and not intentionally.

Zitat von kme

More reports coming, but need to break them down to simplyfied code...

Thanks, all reports are welcome!

klaus.krpano

Zitat von Tuur

kme links don't work
Tuur

For me the links are also not working - 404 error...

klaus.krpano

Zitat von spacerywirtualne

It looks like the ‘setcenter’ in the scrollarea works differently in a 1.22 version

Thanks, but do you have an example for actual testing?
Also note - setcenter() changes the x/y position, so maybe the older version was the wrong working one...

kme

updated the links, it was http vs https

krpano - untitled2

krpano - untitled2

Tuur

.mmm very strange indeed.

When i take out the origin in the 1.22 example the hittest works..

but.. that difference in 'view'..

Tuur

kme

Zitat von Tuur

When i take out the origin in the 1.22 example the hittest works..

true, if origin="0|0|0" it also works as the model is than in the same place as the "hittest" model. but as soon as origin is moved, the hittest model is moved MORE (scale is applied to the model) than the visual model (no scale applied to the offset).

spacerywirtualne

Zitat von klaus.krpano

Thanks, but do you have an example for actual testing?
Also note - setcenter() changes the x/y position, so maybe the older version was the wrong working one...

You can only see this behaviour when you log into my CMS4VR admin panel in the POI editor, but it's not a big problem. I have already corrected it.

klaus.krpano

Zitat von kme

krpano - untitled2
krpano - untitled2

Thanks for the examples!

Zitat von kme

- the VIEW is totally different (scale factor?)

The view is the same, but the preview cube has a different size.
In 1.21 the preview cube was also scaled by depthmap.scale (but that was wrong), in 1.22 not.

In 1.22 there is the new setting image.objectsize for scaling the internal 3D-models of pano, which can be helpful when using fake-3d-transition. Normally this would also scale the preview pano, but especially in this case currently not, because the relation between textured depthmap model and grid-preview-pano is currently somehow undefined. I still need to think and test what might be the best and most logical behavior for the grid-preview-pano here...

Zitat von kme

- the .obj is not moved to the same location as in 1.21 (scale factor not taken into account?)
- hittest is not accurate in 1.22: it is not on where the object is, but it does cause a hittest on where the object would be if it was in the same location as in 1.21 (move mouse somewhere below the cube and it will hit)

This is the same bug, internally the scale and translation matrices had the wrong order and so the origin offsets were not scaled. This will be fixed in 1.22.1.

Best regards,
Klaus

klaus.krpano

Hi,

there is new bugfix release: krpano 1.22.1

It should fix many of the reported problems:

Fix: WebGL 3D-polyline hotspots had a wrong position offset. This was only visible at high-zooms, e.g. in flat-panos (report).
Fix: The WebGL non-borderzoom 3D-polyline hit-testing was inaccurate at high-zooms.
Fix: Custom encryption keys didn't work if the key ID contained a 'z' character (report).
Fix: Registering the license code did not work if the code had two '=' characters at the end and a line-break or white-space in between them.
Fix: Editable textfields (with editenterkey=stop or shiftnewline) didn't lose the input-focus on iOS when the Enter-key was pressed (report).
Fix: Improved String handling in expressions. String operations in expressions could cause that sub-expressions didn't work as expected (report).
Fix: Depending on the actual loading timing, there could be false events (onxmlcomplete, onnewscene) when another pano/scene get loaded while exactly these events are being processed. And this could have several strange/unexpected effects (report).
Fix: Layers with safearea=false and a safearea display, got wrong positions when using changealign() or tweening between relative (percent or negative values) and absolute sizes.
Fix: Improve compatibility by calling the lookto() ondone callback also when the nonblocking setting is not explicitly set to true. Now the lookto is automatically nonblocking when ondone is set and the nonblocking setting not explicitly set to false (report).
Fix: Improve compatibility - when a layer is defined within another layer, its parent setting will be ignored now (report).
Fix: The depthmap.origin position was wrong when depthmap.scale was used (report).

Additionally there is also a small Depthmap/ThreeJS example:

New: ThreeJS: Tesla Example

Best regards,
Klaus

luo jian chao

PanoramicWEB

Hi WebXR edition of 3D version has a parallax problem with oculus quest examples folder samples. Do you have Quest 3 to test?

https://krpano.com/releases/1.22/…eejs_envmap.xml like this sample head movement is must be calibrated.

klaus.krpano

Hi,

that's because the position-tracking isn't enabled automatically:
krpano.com - Plugins - WebVR

The default "auto" mode, gets automatically only enabled when using depthmap-panos, which is not the case there.
That means it need to be enabled manually:

Code

<webvr postracking="true" />

Several ThreeJS examples are affected by this.
These examples will get updated accordingly with the next release.

Best regards,
Klaus

kme

The new 1.22.1 does solve a few earlier reported problems with hittesting.

But I found another one This time with "screentodepth" for threejs hotspots

Exapmples:

https://krpano.kri-soft.be/playground/hittest2/index.html (krpano 1.21)

https://krpano.kri-soft.be/playground/hittest2/index_122.html (krpano 1.22 + threejs)

In the 1.21 example, hittest works as expected.

In the 1.22 example, hittest does not work on the object itself, but it does register a hit when one moves the mouse to the left/under of the motorcycle.

Imgur

Discover the magic of the internet at Imgur, a community powered entertainment destination. Lift your spirits with funny jokes, trending memes, entertaining…

i.imgur.com

Also a question on positioning of the threejs objects.

To make the motorcycles appear in the EXACT same location, I had to map the origin values of krpano obj to tx/ty/tz for threejs object.

Mapping of the tz was origin.z*scale

But mapping of ty ... i had to find manually as no formulate I found was doing an correct mapping (from origin.y =.5 to threejs.y=-54)

Any hints on correct mapping?

Tuur

Hi,

i'm fully unsure, as i didn't tried this yet, but has this to do something with it?
krpano.com - Plugins - ThreeJS
(centermodel)

edit:

eh.. perhaps i mean updateboundingbox.. but it is probably the 'same' or a related bug as in your other post.
Tuur

Jetzt mitmachen!