Hi, Thinking about putting together a test suite for fontspec. Have any of you actually tried the unicode-math test suite? It works by comparing PNG renders of PDF output by xelatex. The main problem with it is false negatives, which happen periodically, and very long execution times. I've heard netpbm is faster than `convert`, but I couldn't get it set up on my machine last I tried.
I'm wondering if such a suite is useable by more people than just the single person who first created the test file. Or if the general idea is sound but the png files should only be created for each user running the test suite. (I.e., first you initialise the tests, and then make changes the files, and then re-run the tests again. You won't find if your files end up the same as mine, but you will be able to see when your changes affected the tests.) Any thoughts? -- Will
