On 01.08.2014 03:22, Carl Worth wrote: > Strictly speaking, the GLSL specification only allows for the '#' to appear in > a shader in two places: > > 1. To introduce a pre-processing directive > > In this case, the '#' must be the first character on a line after > ignoring any comments and horizontal whitespace. > > 2. As part of the token-pasting operator > > In this case, the '#' must appear as part of the sequence "##" and > can only appear in the replacement list of a macro definition. > > Since commit f062f0506a5b827667b7eb52136d8420b7e8113b, glcpp has been strict > about '#' characters, such that any occurence outside of either of the above > two cases would result in a syntax error of "Unexpected HASH_TOKEN" from the > grammar. > > (Prior to that commit, these illegal '#' would be silently ignored and printed > to stdout. And if it hadn't been for that commit, these illegal '#' would have > triggered internal errors with later commits.) > > With this commit, glcpp relaxes its strictness and treats such appearances of > '#' as just another piece of punctuation. This would make it legal to do > something like: > > #define FOO # something here > > and then use "#ifdef FOO" and similar, (though if FOO were ever directly > expanded, this should trigger a syntax error in the main GLSL compiler). > > Relaxing strict conformance compared to GLSL seems somewhat dubious here, but > might be justified for a few reasons: > > 1. GCC's preprocessor allows this, (though C is not GLSL, of course) > > 2. glcpp is generally soft about illegal characters, (compare to > non-ASCII characters which glcpp happily passes through, trusting > the GLSL compiler to catch them). > > 3. At least one game in the wild (Reactive Quake) has been observed
It's 'Reaction Quake'. > with a bogus '#' in it: > > #define USE_DISCARD#line 0 > > Note that the shader (that we've seen) with this line does not ever > *use* the USE_DISCARD macro. Note also that this shader is accepted > by both Nvidia's OpenGL implementation as well as Mesa prior to > the commit referenced above. This change makes the game work again. I'm afraid I don't care about the game enough to report the problem to its developers. :) Tested-by: Michel Dänzer <michel.daen...@amd.com> -- Earthling Michel Dänzer | http://www.amd.com Libre software enthusiast | Mesa and X developer _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/mesa-dev
