Warning fixes #144
Warning fixes #144
Conversation
The attribute only applies crate-wide.
Typical changes: * No need to create a local early in the function, let alone populate it with a value when it will be populted with a value later anyway * No need to carry the destructure output (worse, known-to-be-invalid because it's a field that was previously unused) around as mut; instead, it is assigned to an underscore variable (and dropped immediately because there is no Drop implementation; otherwise we'd have to assign to _ which is hard to read in tuple structs), and the local is created when it is computed.
Conflicts around edhoc.rs were resolved by going with the main version, the fixes will be applied guided by `cargo check` later.
|
Now ready to review (although best reviewed after #142); one warning is left (edhoc.rs line 432 about the kid never being assigned), but I suppose that's best addressed together with adding actual flexibility w/rt the peer's chosen credentials. |
… explicit length This fixes warnings that were legitimately raised about return values ignored, because due to the different mechanisms of length transports, the tests accessed different properties (the returned values) than the implementation used (the old lengths). This also implifies tests a lot, because there is way less copying around between buffer types.
chrysn
force-pushed
the
warning-fixes
branch
from
fe2e4c3 to
ec1b846
Compare
lib/src/edhoc.rs
Outdated
| @@ -45,7 +45,7 @@ pub fn edhoc_key_update( | |||
| _th_3, | |||
| ) = state; | |||
|
|
|||
| let mut prk_new_buf: BytesMaxBuffer = [0x00; MAX_BUFFER_LEN]; | |||
| let mut prk_new_buf: BytesMaxBuffer; | |||
There was a problem hiding this comment.
This one we may have to revert, as hax does not like non-initialized variables. But it could be done in another moment.
There was a problem hiding this comment.
I think the preferred fix here is to not have a single buf but two bufs, not mutate them, and rely on the compiler to see that those two are not used at the same time and thus don't have to sit next to each other on the stack. Proposing the new fix (forcpushibly replacing the merge, which did both a merge and a fix at the same time)
The compiler can see that neither buffer implements a drop, they are not used concurrently, and thus do not need separate space on the stack.
chrysn
force-pushed
the
warning-fixes
branch
from
bcd6925 to
c0f2df8
Compare
Based on #142, this removes the
#, and applies a lot of small refactoring fixes that would probably just have been caught at write time / when they were earlier refactored into the current state.