[cairo] Bad reference images for build test suite

Bryce W. Harrington b.harrington at samsung.com
Wed Jun 26 12:43:14 PDT 2013

Last week psychon mentioned some doubts about the correctness of reference images in the test suite, so I've been poking through them the past few days.  I'm still coming up to speed so it's tough to identify the problems, but the pdf references for the huge-radial test are clearly bogus:


A patch to delete the two bad images is here:


(Linking rather than attaching because the patch is big, and just comprises encoded binary gunk so isn't really reviewable.  Please let me know I should rather be posting these to the list.)

I've written up a post with my understanding of how the reference images work in the test suite; I've glossed some details for brevity, but please correct me on any points I've misunderstood:


Along with this, I'm also working on a tool to identify redundancies and discrepancies amongst the reference images, which I'll post soonish.  I've been trying to use it to spot test cases which would un-break if we removed certain reference images, but haven't yet had much luck:  Some reference images differ within the pdiff threshold and still PASS either way; others (like the above) look different but are still FAILing anyway.

Bryce Harrington
Senior Open Source Developer  -  b.harrington at samsung.com
Open Source Group             -  Samsung Research America

More information about the cairo mailing list