On Tue, Oct 6, 2015 at 9:46 AM, Ben Wagner <[email protected]> wrote:
> I am interested in gx variation fonts being fuzzed. I've created an simple > SIL licensed example font of this type, which can be found at > https://skia.googlesource.com/skia/+/master/resources/fonts/Distortable.ttf > , which is currently being used for testing Skia. Hopefully this can be a > useful starting place for testing these sorts of fonts. > Thanks! I've inserted this file into the corpus, and it added 2 more control flow edges (out of ~15000). Not much, but not zero. The current target function handles any kind of font, so it may be rather slow in finding problems specific to a single font type. For that we better write a separate target function. Would you like to help? --kcc > On Oct 5, 2015 4:28 PM, "Kostya Serebryany" <[email protected]> wrote: > >> >> >> On Mon, Oct 5, 2015 at 7:22 AM, Alexei Podtelezhnikov <[email protected] >> > wrote: >> >>> On Sun, Oct 4, 2015 at 2:40 PM, Kostya Serebryany <[email protected]> >>> wrote: >>> > The data shows that ~500 out of ~1000 functions were never executed. >>> > ... >>> > If we ever get close to ~90% function coverage... >>> >>> About 50% coverage actually sounds reasonable for a given font format. >>> For example, Freetype has 3 hinters for different fonts. Of course, we >>> do not use them together. It took a while before the Fuzzer >>> "discovered" CFF space. >> >> >> About two weeks on 20 CPU cores, to be precise. >> if I had a test input that uses CFF, it would have been much quicker. :) >> >> >>> I might be totally off base here, but coverage >>> for a given font format would be more informative. FreeType supports a >>> large but limited number of font formats >>> http://freetype.org/freetype2/docs/index.html . >>> >> >> How do you identify that a given file belongs to one format or another? >> Some magic in the beginning? >> We can modify the target function to immediately reject all formats but >> one, and we'll get what you are asking for. >> This is where my knowledge about this domain area (fonts) ends and where >> I may need your help. >> >> --kcc >> >> _______________________________________________ >> Freetype-devel mailing list >> [email protected] >> https://lists.nongnu.org/mailman/listinfo/freetype-devel >> >>
_______________________________________________ Freetype-devel mailing list [email protected] https://lists.nongnu.org/mailman/listinfo/freetype-devel
