Collin Funk <[email protected]> writes:

>> There are still a couple of test failures. Namely, what does not work?
>> - conversions from [u]intptr_t to a pointer
>> - <fenv.h> glibc-specific functions
>> - asm with pointer arguments
>> - at-init
>
> Makes sense. My understanding is that it stores capabilities in
> pointers. So I would assume we would need adjustments similar to CHERI.

Oh, the documentation even uses CHERI as an example of the capability
model [1]:

    InvisiCaps can also be thought of as a software implementation of
    CHERI. Unlike CHERI, InvisiCaps are more compatible (pointers are
    64-bit, not 128-bit or 256-bit) and InvisiCaps have a more
    deterministic use-after-free story.

Collin

[1] https://fil-c.org/invisicaps

Reply via email to