We currently don't have a fuzzer for icc profiles to test qcms. It would be nice to have one.
Jesse, I was talking to Jeff today about whether he thinks it would make sense for us to get some fuzzers specific to the graphics code. He said that he thinks that I'm not completely crazy, but he's not sure what such fuzzers would do, and what kind of information he needs to provide for you in order for us to get to an estimate of whether such a fuzzer would be worth having, and finally get one. This is one of the bugs in this series. Feel free to ask for any information that you think might be helpful for getting this type of fuzzer. Thanks!
Already done. It fuzzes ICC profiles inside PNG images.
Is it possible to get access to these fuzzers?
It's part of a bigger framework and doesn't run without shared libraries. What it basicaly does is using an ICC profile, decompressing it, mutating it, compressing it and put the malformed profile in the iCCP ancillary chunk.
There's a bunch of new icc parsing code that's landed. It would be good to refuzz this stuff especially with the gfx.color_management.enablev4 pref set to true.
Is it possible to look into getting some fuzzing done. I've looking at a few performance problems then we should be good to turn this on.
Yes, we definitely need to look at it again.
From our IRC discussion I'm going to prepare a set of PNG test cases including some images with ICCv4 profiles.