[Piglit] FW: [PATCH] fixed oes compressed etc2 texture miptree failure

Brian Paul brianp at vmware.com
Tue Apr 29 07:05:12 PDT 2014


The content of your original patch looks OK, but it should probably be 
split into two patches: one for the ktx change, and another for the 
viewport fix.

Can you please test changing the default window size to 160x160 and 
rerun on Win8 to verify?  I'll do the same here on Linux.

-Brian

On 04/29/2014 05:03 AM, Guo, Johney wrote:
> Bumping up to 160 is simple way to fix the "warn"s on Win8.
> For further compatibility, it is better to set the default window fix sized to disable resize function.
> BTW,Will my patch be accepted for this test?
>
> On 04/27/2014 12:45 PM, Ian Romanick wrote:
>> On 04/24/2014 03:41 AM, Guo, Johney wrote:
>>> See tests/util/piglit-framework-gl/piglit_glut_framework.c:84:
>>> default_reshape_func(int w, int h)
>>> {
>>> 	if (piglit_automatic &&
>>> 	    (w != piglit_width ||
>>> 	     h != piglit_height)) {
>>> 		printf("Got spurious window resize in automatic run "
>>> 		       "(%d,%d to %d,%d)\n", piglit_width, piglit_height, w, h);
>>> 		piglit_report_result(PIGLIT_WARN);
>>> 	}
>>>
>>> 	piglit_width = w;
>>> 	piglit_height = h;
>>>
>>> 	glViewport(0, 0, w, h);
>>> }
>>>
>>> My OS is win8. If piglit_width < 160, window manager will force it to 160.
>>
>> There was a similar issue a few years ago with Win7.  I believe Brian
>> Paul fixed this by modifying the framework to enforce a minimum window
>> size depending on the operating system.  It's better to fix this issue
>> once for all tests than change each individual test.  This is
>> especially true since we may add more tests with too-small window
>> sizes (since we're not on Windows).
>>
>> Brian, does that sound right?
>
> Basically, I changed the default window size to be 150x150 because on Windows (7 and maybe XP) if the window size was less than 116 pixels wide, it would get bumped up to 116 and that would cause the test to fail.
>
> I did a quick search of git history and found this one:
>
> commit 1894ef34c5e3cc0daefd1241b31a2b2ae32c3347
> Author: Brian Paul <brianp at vmware.com>
> Date:   Fri Apr 5 12:32:46 2013 -0600
>
>       lodbias: increase window size to fix test on Windows
>
>       Before, the 96x96 window got bumped up to 116x96 by Windows and we
>       saw a "Got spurious window resize in automatic run" message and the
>       test reported "warn".
>
> So the rule of thumb now is don't set the config.window_width, height vars unless you really have to, and be prepared for window resizes if you do.
>
> I've rarely tested Windows 8 so I wasn't aware of any problems there with the window size.  If width=160 is the new magic number, we may have to bump up config.window_width, height again.
>
> Johney, can you try bumping up the default window size in tests/util/piglit-framework-gl.h and see if that solve the problem?
>
> I can likewise do that here and see if there's any unexpected regressions from changing the window size.
>
>
>>
>>> Many piglit tests have set configure width as 150,  then they will all report as "warn".
>>> At least the following,
>>> $ grep -Irne "\<150\>" tests | grep width
>>> tests/fbo/fbo-blit-d24s8.c:45:  config.window_width = 150;
>>> tests/fbo/fbo-blit.c:43:        config.window_width = 150;
>>> tests/fbo/fbo-copypix.c:40:     config.window_width = 150;
>>> tests/fbo/fbo-readdrawpix.c:40: config.window_width = 150;
>>> tests/spec/arb_es2_compatibility/arb_es2_compatibility-depthrangef.c:36:        config.window_width = 150;
>>> tests/texturing/incomplete-texture.c:38:        config.window_width = 150;
>>> tests/texturing/shaders/textureSize.c:55:       config.window_width = 150;
>>> tests/texturing/texsubimage.c:42:       config.window_width = 150;
>>> tests/util/piglit-framework-gl.h:272:                config.window_width = 150;
>>>
>>> What is more, 	glViewport(0, 0, w, h)  will change the viewport to (w,h),  so in each glut display() function, glViewPort() should be recalled for validity
>>
>
>
> The last time I bumped up the default window size I had to fix a bunch
> of tests that had hard-coded dimensions.  Looks like there's a few more
> to tend to...
>
> -Brian
>



More information about the Piglit mailing list