Vousk-prod.

Members
  • Content count

    856
  • Joined

  • Last visited

  • Days Won

    5

Vousk-prod. last won the day on December 19 2015

Vousk-prod. had the most liked content!

About Vousk-prod.

  • Rank
    Advanced Member
  • Birthday

Contact Methods

  • Website URL
    http://vousk-prod.net

Profile Information

  • Gender
    Not Telling
  • Location
    Strasbourg - France

Recent Profile Visitors

819 profile views
  1. Hi everyone ! I've just noticed that when the rendering canvas size is quite big, above a certain size the 3D drawing area is suddenly cropped (to half the size of the rendering canvas, it seems). If you lower the width of the canvas, the height limit increases, and vice versa. And a strange fact: critical values are different depending on the context. Here are some cases (on my computer, values may differ on yours) - On a simple web page: 3500px x 1024px : ok 3500px x 1025px (and above) : cropped 3637px x 1024px : ok 3638px (and above) x 1024px : cropped I tried to reproduce that on the PG, and the values are not the same: 3500px x 2368px : ok 3500px x 2369px (and above) : cropped 3536px x 2368px : ok 3537px (and above) x 2368px : cropped The width around 3500 is an arbitrary value, you can put quite more if height is low enough (and vice versa). I don't know if the values will be different on your devices, but you can test that with this PG : http://www.babylonjs-playground.com/#17HUVT#1 Is it expected behavior or a bug ?
  2. @Kesshi Mhm ok. You're right, the colored outline was just a side effect of using a red-but-in-fact-fully-transparent background to demonstrate the Chrome alpha bug.
  3. @Kesshi I know that, I'm using Color4(0, 0, 0, 0) in real cases. In the PG I put red background to be able for others to more easily see the bug. The problem here is not the ability or not to make transparent, colored or semi transparent background, the problem is that the new screenshot method does not correctly take the choosen clearColor in account (in Chrome), whereas the old method did, and a classic canvas to data also is ok.
  4. @Nabroski First, to do whatever you like with a buffer, you first need your buffered image at the right size, Photoshop is perfect (I'm using it everyday) but it cannot do miracles, upscale a picture still pixelates it, even in 2017 . Secondly, yes as a developper I can code my own screnshot function (in fact, it's already the case, in my app I'm using my own function, but it dynamically changes the webpage layout, which is not something we can consider in BJS, we need to provide a screenshot function that do not alter the HTML). The idea of my post here is to chase the bugs, for everyone, not only for me, I can get around problems, but this is not the main purpose of a library, the idea to provide a functionality is to avoid people to have to write patches to use it.
  5. @Nabroski There are 3 major problems with your solution : 1. you cannot specify custom render size, 2. the usage is completely different from the previous screenshot process (with options to specifiy precision, widht and/or height, etc), 3. it would be a bit odd to provide a screenshot function to users and to tell them "it's not working, you need to dump yourself the frame buffer"
  6. @Nabroski Yup, but the old render target method works. And a "manual" canvas dump ( canvas.toDataURL("image/png") ) also works with no problem in Chrome. Also the fact that the final screenshot is upscaled tends to show that there is something strange in the process involved in new screenshot function.
  7. Yes, I still repro it. With the PG above, however scene clearColor alpha is 0, with Chrome the screenshot has the background color instead of transparent. And with Firefox (classic and developper versions), some objects edges are blueish.[EDIT: not a bug, to avoid outlines artifacts, use premultiplied alpha for clearColor, see Kesshi's post page 2] I also noticed that when choosing a render width a lot greater than canvas original size, the image is pixelated (it looks like the render happened at canvas original size and the produced image is upscaled afterwards). You can check that with the same PG, using width:4096 instead of 1024.
  8. Do you have any idea, DK, concerning this bug ?
  9. Great news, but I think there is no need to rush, since all browsers will (as usual) take time to fully and natively support WebGL 2.0, probably still a full year... But I can be wrong, that would be awesome ! How is Edge roadmap on that ?
  10. Nice !
  11. I'm on mobile right now, I don't have the full gui in front of my eyes, but if I remember well, just edit the first post, then in "advanced edition" (or something similar) you'll be able to modify the title.
  12. To mark post as solved, simply edit the subject with [Solved]. We don't have a button for that in this forum unfortunatly.
  13. I don't clearly get what you're talking about, but when I click an hex tile, some strange things definetly are happening. I don't know what I should see, but sometimes a click changes material for a tile, sometimes it makes all "flat" tiles invisible and "layered" tiles appears... But in console I see that some .mtl are missing, couldn't it be the problem ? (when a texture or a material is not found, WebGL render the object with "nothing" as material, so the object is just invisible).
  14. Hmmm, sorry to bother again @Deltakosh but when scene's clearColor has alpha, the PNG generated by the new Screenshot function does not take alpha in account in Chrome (it works in FF). And also (but in FF only this time ) there is a strange colored artifact happening on some objects edge and lineMesh. [EDIT: not a bug, to avoid outlines artifacts, use premultiplied alpha for clearColor, see Kesshi's post page 2] PG : http://www.babylonjs-playground.com/#QNSHJ#9