Re: [FreeWRL-develop] max vertical window size 1024px? (just Linux?)

[Greg R. <ne...@po...>]

Ian wrote:

> On 25/11/14 10:24 AM, Ian Stakenvicius, Aerobiology wrote:
>> On 25/11/14 09:55 AM, John Alexander Stewart wrote:
>>> Hi Greg;
>>>
>>> I don't believe that there are limits on the window size - on OSX, it
>>> can be much larger than 1024 pixels! But the OSX code is different than
>>> Linux/Windows, of course.
>>>
>>> I don't currently have a Linux machine configured, so can't verify.
>>> Anyone else?
>>>
>>> JohnS.
>>
>> I'm looking into it now.  I too, though, don't recall running into this
>> issue.

> Confirmed -- I do not see any such resizing issues on linux, with
> current master.  Everything renders as expected whether it be maximized
> or resized in all sorts of odd sizes, on my 1600x1440 desktop.

Well, that's very weird.  I compiled from scratch on a new machine just
a few days ago (both old svn and current git codebases), and it has exactly
the same problem.  Chromium/WebGL on the same machine, on the other hand,
has no such limits.  (I can't test Chrome on the older machine; my OpenGL
drivers are too old/incapable.)

Christoph Willing wrote:

> In another project altogether, we had some users with a similar problem. 
> It turned out that it was due to a limitation of the graphics card; 
> specifically the maximum size of the GL viewport - any window larger 
> than the max would show grey or black in the additional window area. 
> Modern graphics have much bigger viewports so you'd never see it any 
> more. You can check the viewport size from a terminal by running:
>      glxinfo -l |grep DIM
> 
> e.g.
> chris@d6:~$ glxinfo -l |grep DIM
>      GL_MAX_VIEWPORT_DIMS = 16384, 16384
>      GL_MAX_VIEWPORT_DIMS = 16384, 16384

I could easily believe that on the older machine; it's an Atom CPU with
some old Intel integrated graphics chipset (82945G/GZ), but in fact:

% glxinfo -l | grep DIM
    GL_MAX_VIEWPORT_DIMS = 4096, 4096

...so it seems that it _should_ be OK.  OTOH:

% glxinfo -l | grep 1024
        GL_MAX_PROGRAM_PARAMETERS_ARB = 1024
        GL_MAX_PROGRAM_LOCAL_PARAMETERS_ARB = 1024
        GL_MAX_PROGRAM_LOCAL_PARAMETERS_ARB = 1024

Those sound like argument counts or something, though, not pixel limits.

On the new machine (NUC, so still Intel integrated graphics, but HD 5000
instead of 945):

% glxinfo -l | grep DIM
    GL_MAX_VIEWPORT_DIMS = 16384, 16384
    GL_MAX_VIEWPORT_DIMS = 16384, 16384

% glxinfo -l | grep 1024
        GL_MAX_PROGRAM_NATIVE_PARAMETERS_ARB = 1024
        GL_MAX_PROGRAM_NATIVE_INSTRUCTIONS_ARB = 1024
        GL_MAX_PROGRAM_NATIVE_PARAMETERS_ARB = 1024
        GL_MAX_PROGRAM_NATIVE_ALU_INSTRUCTIONS_ARB = 1024
        GL_MAX_PROGRAM_NATIVE_TEX_INSTRUCTIONS_ARB = 1024
        GL_MAX_PROGRAM_NATIVE_TEX_INDIRECTIONS_ARB = 1024

Anyway, thanks for the sanity checks and suggestions.  Clearly it's somehow
local to my setup (maybe how I compile Freewrl?), so I'll dig further.

Greg

P.S.  three.js includes a VRML loader; has anyone played with that?  I
      assume it handles only geometry/materials, not behaviors, but I
      haven't had a chance to try it yet.