[Piglit] [PATCH] Add test glsl-uniform-out-of-bounds-2.c

[Frank Henigman]

On Mon, Nov 26, 2012 at 2:27 PM, Eric Anholt <eric at anholt.net> wrote:

> Frank Henigman <fjhenigman at google.com> writes:
> > diff --git a/tests/all.tests b/tests/all.tests
> > index 36c5847..2cf1ef4 100644
> > --- a/tests/all.tests
> > +++ b/tests/all.tests
> > @@ -240,6 +240,7 @@ add_plain_test(shaders, 'glsl-novertexdata')
> >  add_plain_test(shaders, 'glsl-preprocessor-comments')
> >  add_plain_test(shaders, 'glsl-reload-source')
> >  add_plain_test(shaders, 'glsl-uniform-out-of-bounds')
> > +add_plain_test(shaders, 'glsl-uniform-out-of-bounds-2')
>
> I think this could safely be a concurrent test.
>
> > +static GLint prog;
>
> Looks like this ought to live next to vs, fs declarations.
>
> > +void
> > +piglit_init(int argc, char **argv)
> > +{
> > +     GLint vs, fs;
> > +     int i, j, k;
> > +     bool pass = true;
> > +     GLint numActiveUniform;
>
> Generally we try to go for names_with_underscores.
>
> > +     // for each array in shader
> > +     for (k = 0; k < numActiveUniform; ++k) {
> > +             GLchar name[99];
> > +             GLint numActiveElements;
> > +             GLenum type;
> > +             int size;
> > +             glGetActiveUniform(prog, k, ARRAY_SIZE(name), NULL,
> > +                                                &numActiveElements,
> &type, name);
> > +             if (name[1] != 0)
> > +                     continue;
> > +             if (name[0] == 'v')
> > +                     size = 4;
> > +             else if (name[0] == 'm')
> > +                     size = 16;
> > +             else
> > +                     continue;
> > +             printf("array '%s' active elements %d\n", name,
> numActiveElements);
> > +
> > +             // for each index in array, plus some before and after
> > +             for (i = -2; i < 6; ++i) {
> > +                     bool isActive = 0 <= i && i < numActiveElements;
> > +                     GLchar element[9];
> > +                     GLint loc;
> > +                     sprintf(element, "%s[%d]", name, i);
> > +                     loc = glGetUniformLocation(prog, element);
> > +
> > +                     // check result of glGetUniformLocation
> > +                     if (loc == -1) {
> > +                             if (isActive) {
> > +                                     printf("FAIL: no location for
> active %s\n", element);
> > +                                     pass = false;
> > +                             }
> > +                     } else {
> > +                             if (!isActive) {
> > +                                     printf("FAIL: got location for
> inactive %s\n", element);
> > +                                     pass = false;
> > +                             }
> > +
> > +                             // write location
> > +                             if (size == 4) {
> > +                                     glUniform4fv(loc, 1, data);
> > +                             } else {
> > +                                     glUniformMatrix4fv(loc, 1,
> GL_FALSE, data);
> > +                             }
> > +                             // read back
> > +                             for (j = 0; j < size; ++j)
> > +                                     v[j] = 0;
> > +                             glGetUniformfv(prog, loc, v);
> > +                             // compare
> > +                             for (j = 0; j < size; ++j) {
> > +                                     if (v[j] != data[j]) {
> > +                                             printf("FAIL: wrong value
> in %s\n", element);
> > +                                             pass = false;
> > +                                             break;
> > +                                     }
> > +                             }
>
> Being able to round-trip data through the various components of a
> uniform array/matrix seems like a very separate thing to test from
> glsl-uniform-out-of-bounds-2.  This test would make a lot more sense to
> me without it.  In that case, this loop would just be checking that
> out-of-bounds stuff gets -1 location (a good thing to test), and
> computing the bounds of the valid locations.
>

Mesa currently will return a location for elements past the point where
glGetActiveUniform says they are active (I posted a patch for that but
was asked for a test, hence this thread) and I wanted to see if those
locations worked, as an indication of whether glGetActiveUniform or
glGetUniformLocation was wrong.  Would it be ok if I only test the
round trip on those suspect locations?   Thanks.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/piglit/attachments/20121126/785a60bb/attachment.html>