[Piglit] [PATCH 2/2] gl-1.0-readpixsanity: Eliminate use of variable length arrays.

[Jan Vesely]

On Wed, 2014-10-29 at 11:48 +0000, jfonseca at vmware.com wrote:
> From: José Fonseca <jfonseca at vmware.com>
> 
> Non-standard GCC extension.

just FYI, it's not GCC extension, it's c99 standard.
maybe we can use more msvc compatible set of gcc flags (-std=c89
-pedantic? ) for code that needs to compile in windows

jan

> ---
>  tests/spec/gl-1.0/readpix.c | 64 ++++++++++++++++++++++++++-------------------
>  1 file changed, 37 insertions(+), 27 deletions(-)
> 
> diff --git a/tests/spec/gl-1.0/readpix.c b/tests/spec/gl-1.0/readpix.c
> index 7ee0be9..26f44a6 100644
> --- a/tests/spec/gl-1.0/readpix.c
> +++ b/tests/spec/gl-1.0/readpix.c
> @@ -100,7 +100,9 @@ check_rgba(void)
>  	double err;
>  	int xerr, yerr;
>  	float expected[4], expected_rgba[4], actual_rgba[4];
> -	GLfloat buf[piglit_width][piglit_height][4];
> +	const int w = piglit_width;
> +	const int h = piglit_height;
> +	GLfloat *buf = (GLfloat *)malloc(h * w * 4 * sizeof *buf);
>  	GLfloat dr, dg, db, da;
>  	GLint rbits, gbits, bbits, abits;
>  	glGetIntegerv(GL_RED_BITS, &rbits);
> @@ -119,19 +121,18 @@ check_rgba(void)
>  		glClear(GL_COLOR_BUFFER_BIT);
>  
>  		/* Read the buffer: */
> -		glReadPixels(0, 0, piglit_width,
> -			piglit_height, GL_RGBA, GL_FLOAT, buf);
> +		glReadPixels(0, 0, w, h, GL_RGBA, GL_FLOAT, buf);
>  
>  		/*
>  		 * Now compute the error for each pixel, and record the
>  		 * worst one we find:
>  		 */
> -		for (y = 0; y < piglit_height; ++y) {
> -			for (x = 0; x < piglit_width; ++x) {
> -				dr = fabs(buf[y][x][0] - expected[0]);
> -				dg = fabs(buf[y][x][1] - expected[1]);
> -				db = fabs(buf[y][x][2] - expected[2]);
> -				da = fabs(buf[y][x][3] - expected[3]);
> +		for (y = 0; y < h; ++y) {
> +			for (x = 0; x < w; ++x) {
> +				dr = fabs(buf[y*w*4 + x*4 + 0] - expected[0]);
> +				dg = fabs(buf[y*w*4 + x*4 + 1] - expected[1]);
> +				db = fabs(buf[y*w*4 + x*4 + 2] - expected[2]);
> +				da = fabs(buf[y*w*4 + x*4 + 3] - expected[3]);
>  				err =
>  				    fmax(error_bits(dr, rbits),
>  				    fmax(error_bits(dg, gbits),
> @@ -156,7 +157,7 @@ check_rgba(void)
>  						expected_rgba[j] = 
>  							expected[j];
>  						actual_rgba[j] = 
> -							buf[y][x][j];
> +							buf[y*w*4 + x*4 + j];
>  					}
>  				}
>  			}
> @@ -187,6 +188,8 @@ check_rgba(void)
>  			
>  	}
>  
> +	free(buf);
> +
>  	return pass;
>  } /* check_rgba */
>  
> @@ -196,8 +199,10 @@ check_depth(void)
>  	int i, x, y;
>  	int thresh = 1;
>  	bool pass = true;
> -	GLdouble expected, expected_depth, actual_depth;
> -	GLuint buf[piglit_width][piglit_height];
> +	GLdouble expected, expected_depth, actual, actual_depth;
> +	const int w = piglit_width;
> +	const int h = piglit_height;
> +	GLuint *buf = (GLuint *)malloc(h * w * sizeof *buf);
>  	double current_error = 0.0;
>  	GLfloat dd;
>  	double err;
> @@ -219,25 +224,25 @@ check_depth(void)
>  		 * than a GLfloat.  Since this is just a sanity check, we'll
>  		 * use integer readback and settle for 32 bits at best.
>  		 */
> -		glReadPixels(0, 0, piglit_width,
> -			piglit_height, GL_DEPTH_COMPONENT,
> +		glReadPixels(0, 0, w, h,
> +			GL_DEPTH_COMPONENT,
>  			GL_UNSIGNED_INT, buf);
>  
>  		/*
>  		 * Now compute the error for each pixel, and record the
>  		 * worst one we find:
>  		 */
> -		for (y = 0; y < piglit_height; ++y) {
> -			for (x = 0; x < piglit_width; ++x) {
> -				dd = abs(buf[y][x]/4294967295.0
> -					- expected);
> +		for (y = 0; y < h; ++y) {
> +			for (x = 0; x < w; ++x) {
> +				actual = buf[y*w + x]/(double)0xffffffffU;
> +				dd = abs(actual - expected);
>  				err = error_bits(dd, dbits);
>  				if (err > current_error) {
>  					current_error = err;
>  					xerr = x;
>  					yerr = y;
>  					expected_depth = expected;
> -					actual_depth = buf[y][x]/4294967295.0;
> +					actual_depth = actual;
>  				}
>  			}
>  		}
> @@ -258,6 +263,8 @@ check_depth(void)
>  			actual_depth);
>  	}
>  
> +	free(buf);
> +
>  	return pass;
>  } /* check_depth */
>  
> @@ -278,7 +285,9 @@ check_stencil(void)
>  {
>  	int i, x, y;
>  	bool pass = true;
> -	GLuint buf[piglit_width][piglit_height];
> +	const int w = piglit_width;
> +	const int h = piglit_height;
> +	GLuint *buf = (GLuint *)malloc(h * w * sizeof *buf);
>  	GLuint expected;
>  	GLint sbits;
>  	glGetIntegerv(GL_STENCIL_BITS, &sbits);
> @@ -291,14 +300,13 @@ check_stencil(void)
>  		glClear(GL_STENCIL_BUFFER_BIT);
>  		pass &= piglit_check_gl_error(GL_NO_ERROR);
>  
> -		glReadPixels(0, 0, piglit_width,
> -			piglit_height, GL_STENCIL_INDEX,
> -			GL_UNSIGNED_INT, buf);
> +		glReadPixels(0, 0, w, h,
> +			GL_STENCIL_INDEX, GL_UNSIGNED_INT, buf);
>  		pass &= piglit_check_gl_error(GL_NO_ERROR);
>  
> -		for (y = 0; y < piglit_height && pass; ++y) {
> -			for (x = 0; x < piglit_width; ++x) {
> -				if (buf[y][x] != expected) {
> +		for (y = 0; y < h && pass; ++y) {
> +			for (x = 0; x < w; ++x) {
> +				if (buf[y*w + x] != expected) {
>  					pass = false;
>  					break;
>  				}
> @@ -315,9 +323,11 @@ check_stencil(void)
>  		printf("\tStencil failed at (%i, %i).\n",
>  			x, y);
>  		printf("\t\tExpected %i; got %i.\n",
> -			expected, buf[y][x]);
> +			expected, buf[y*w + x]);
>  	}
>  	
> +	free(buf);
> +
>  	return pass;
>  } /* check_stencil */
>  

-- 
Jan Vesely <jan.vesely at rutgers.edu>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: This is a digitally signed message part
URL: <http://lists.freedesktop.org/archives/piglit/attachments/20141029/ed80e327/attachment.sig>