Sign extend signed integers - #238
Conversation
|
Could you add tests to the |
Since there's no change to attributes I think it's alright as is (current test cases were updated) |
|
Is there a way to reduce the duplication of code between signed/non-signed? I'm open to having this as a separate issue. |
|
Friendly ping @korrat are you still interested in pursuing this PR? |
4d118dc to
547d83c
Compare
|
Sorry, @sharksforarms, totally forgot about this one. I'm still interested in pursuing this PR.
I've tried to merge the |
Previously, the two macros to implement DekuRead for integer primitives contained nearly identical code, with the exception of sign-extending in the signed integer implementation. With this commit, signed integers reuse the code of their unsigned counterparts and add sign-extending on top.
|
Here's a proposal to reduce the duplication. Let me know if this is what you had in mind. Alternatively, it should be possible to use a single implementation for the integer types and instead proxy the float implementations through unsigned integers, similarly to how it's done for the signed integers in this commit. |
|
I can reproduce the compile error on master, so I think it is unrelated to the changes in this PR. I think the error is related to rust-lang/rust#79609. Defining |
|
Hey @korrat sorry for my delay. The PR looks good at first glance, I'll take some time this week to look at it. A separate PR would be appreciated! |
|
Thanks! Released in 0.13.0 |
3 participants
This PR changes the implementation of
DekuReadfor signed integer primitive types, so that reads with a bit size less than the size of the underlying type are sign extended.Fixes #236.