Created 2022-04-11, last updated 2022-04-16 18:09:48 UTC
- While working on adding some more unit tests I found some undefined
behavior using Miri. Filed an
- Put up a PR to
partially fix that UB, by avoiding use of
allocate_buffer in the
file info code.
- Found some potential UB in the file info code with
-Zmiri-tag-raw-pointers, put up a
PR to fix it.
- I'd like to enable
-Zmiri-tag-raw-pointers in our Miri tests in
general, but it's catching some issues with
DevicePath. This one is
tricky because unlike most protocols,
DevicePath is really
unsized. But we don't currently implement it that way.
- I've been using Miri more in uefi-rs to test our unsafe code, which we
have quite a lot of. One issue I've had is that I find myself guessing
a lot as to what Miri is trying to tell me. The error messages are
relatively terse, and while a lot has been written about stacked
borrows, it's not clear what the best resource to read is. As a first
step to improving this situation I filed an
issue in the Miri
repo suggesting adding documentation for Miri's output to the
readme. Even just documenting the notation used in the output would be
a good start.