On Sat, Jan 7, 2017 at 5:23 PM, Roland Scheidegger <srol...@vmware.com> wrote: > Am 07.01.2017 um 15:51 schrieb Marek Olšák: >> From: Marek Olšák <marek.ol...@amd.com> >> >> This fixes broken depth texturing after: >> >> commit 22639a6e19f95902aef23474ad672bf489231ea7 >> Author: Timothy Arceri <timothy.arc...@collabora.com> >> Date: Mon Nov 21 00:29:29 2016 +1100 >> >> st/mesa: get Version from gl_program rather than gl_shader_program >> --- >> src/mesa/main/ff_fragment_shader.cpp | 3 ++- >> 1 file changed, 2 insertions(+), 1 deletion(-) >> >> diff --git a/src/mesa/main/ff_fragment_shader.cpp >> b/src/mesa/main/ff_fragment_shader.cpp >> index 48b84e8..a13a5de 100644 >> --- a/src/mesa/main/ff_fragment_shader.cpp >> +++ b/src/mesa/main/ff_fragment_shader.cpp >> @@ -1215,21 +1215,22 @@ create_new_program(struct gl_context *ctx, struct >> state_key *key) >> p.shader_program = _mesa_new_shader_program(0); >> >> /* Tell the linker to ignore the fact that we're building a >> * separate shader, in case we're in a GLES2 context that would >> * normally reject that. The real problem is that we're building a >> * fixed function program in a GLES2 context at all, but that's a >> * big mess to clean up. >> */ >> p.shader_program->SeparateShader = GL_TRUE; >> >> - state->language_version = 130; >> + /* Don't use 1.30, because depth texturing behaves differently there. */ > Could you elaborate on this?
Yeah it wasn't a completely accurate statement. IIRC (off the top of my head), the legacy GLSL shadow functions follow the depth texture mode and return vec4. The GLSL 1.30 shadow functions return float and ignore the depth texture mode. That's a shader and state dependency that's difficult to deal with. st/mesa uses a simple but not completely correct solution: if the shader declares GLSL >= 1.30 and the depth texture mode is GL_ALPHA (000X), it sets the XXXX swizzle instead. Thus, the GLSL 1.30 shadow function will get the result in .x and legacy shadow functions will get it in .w as expected. It's an acceptable deviation from the spec considering how rare the GL_ALPHA setting is, but one piglit test checking GL_ALPHA showed an obvious failure there. Marek _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/mesa-dev
