On 26 August 2013 16:53, Ian Romanick <[email protected]> wrote: > On 08/26/2013 04:14 PM, Eric Anholt wrote: > >> Ian Romanick <[email protected]> writes: >> >> From: Ian Romanick <[email protected]> >>> >>> Instead of replicating the same code over and over and over, just put it >>> in one file that each shader test can reference. >>> >> >> I've been meaning to kill "[vertex/fragment shader file]" support from >> shader_runner for a while. :( >> >> What I don't like about it is that it scatters the information about a >> particular test to more places, which shader_runner had helped reduce >> significantly. And this feature is used very rarely -- looking at the >> current list of tests using it, they mostly look like the first few >> shader_runner tests we made. >> >> For the changes you wanted to do in this series, a sed job looks like it >> would have been just fine. >> > > I think it's not used much because most people don't know it's available. > These are the (only?) cases where it's useful: you don't want to type the > dummy, pass-through vertex shader over and over and over. Maybe we should > replace it with a more limited "[vertex shader passthrough]" option? >
FWIW the reason I don't use it isn't because I don't know it's available. It's because when writing a test, it feels easier for me to just pound out the dummy pass-through vertex shader rather than to go hunt around nearby directories to find the shader I want. I also find it slightly easier to code review others' shader tests if they don't use this feature, because it means I can completely understand the shader test in isolation, and I don't feel compelled to double-check that the file they're referring to actually exists (and does what it claims). I like the idea of a "[vertex shader passthrough]" option--it would neatly address both of the reasons why I don't currently use [vertex shader file]. However, I don't have terribly strong opinions about this stuff. If you want to just leave the series as is, that's fine by me as well.
_______________________________________________ Piglit mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/piglit
