Stream: wasmtime

Topic: Debugging info fuzzing


view this post on Zulip Till Schneidereit (Dec 03 2020 at 14:39):

This is an interesting paper on testing debug info transforms: https://arxiv.org/abs/2011.13994

I wonder if any of this applies to the Wasmtime/Cranelift DWARF support? /cc @Yury Delendik

view this post on Zulip Yury Delendik (Dec 03 2020 at 14:44):

It applies, that's why we prefer gimli over libdwarf.

view this post on Zulip Till Schneidereit (Dec 03 2020 at 14:45):

heh, good point :smile:

view this post on Zulip Yury Delendik (Dec 03 2020 at 14:46):

And that's why explicitly using -g for processing DWARF info is somewhat a good idea

view this post on Zulip Yury Delendik (Dec 03 2020 at 14:48):

in theory, wasmtime shall act like a sanitizer of DWARF information, but it is not proven

view this post on Zulip Till Schneidereit (Dec 03 2020 at 15:09):

I wonder if it'd be feasible to hook it up to the framework described in the paper?

view this post on Zulip Chris Fallin (Dec 03 2020 at 17:22):

Thanks for this link, I'll read the paper later! After reading the abstract it definitely sounds to be in the same family as the regalloc checker (dataflow equivalence verification); maybe there's a way that we can add another fuzz oracle inspired by this

view this post on Zulip Till Schneidereit (Dec 03 2020 at 17:38):

yeah, I was thinking something along those lines. I don't think it's something that should be a particular priority, but it seemed interesting :smile:


Last updated: Dec 23 2024 at 13:07 UTC