On 3 August 2017 at 19:24, Grigori Goronzy <g...@chown.ath.cx> wrote:
>> I don't know if we need the no_error here. The test should check if >> things work correctly when it's set. >> Everything else falls outside the scope of the extension. >> > > I added testing with no-error==false because there were some issues with the > EGL implementation of the no-error extension, which then switched on > no-error mode if any context flag was set. Some other tests *might* also > catch such issues, but here's the obvious place. Anyway, that's the > motivation for doing these "control tests". What do you think? > Hmm good call. If the driver is bugged and sets no_error w/o the user knowledge, at the same time the latter expects glGetError to work, things will go horribly bad. At the same time, let's drop the no_error check around glGetIntegerv(GL_CONTEXT_FLAGS, &context_flags). >>> + attribs[ai++] = EGL_CONTEXT_FLAGS_KHR; >>> + attribs[ai++] = flags; >>> + } >>> + /* Always use OpenGL 2.0 or OpenGL ES 2.0 to keep this test >>> reasonably >>> + * simple; there are enough variants as-is. >> >> Another hint - GL_KHR_no_error requires OpenGL{,ES} 2.0 >> > > OK, I'll add a check to validate that the implementation doesn't accept > no-error for GL/ES 1.x. > Don't think that's needed here. That is an implementation detail of the GL extension. I was merely suggesting to add a comment. -Emil _______________________________________________ Piglit mailing list Piglit@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/piglit