Stream: t-lang/wg-unsafe-code-guidelines

Topic: packed UB error


gnzlbg (Nov 29 2018 at 08:43, on Zulip):

@RalfJ I read your comment https://github.com/rust-lang/rfcs/pull/2582#discussion_r237383311 but it is still not clear to me.

The RFC adds a migration path for people using repr(packed) today, does the RFC also turns this warning into an error?

I think that when the hole thing is over we should write a post-mortem about repr(packed) because what has happened is pretty much an example of how not to deal with an I-Unsound bug IMO.

RalfJ (Nov 29 2018 at 08:57, on Zulip):

does the RFC also turns this warning into an error?

No, I wanted to keep the RFC focused on the "create raw ref" operation. But if people think that makes more sense, I'd be happy to extend it to also finalize our story around errors/lints for references to packed fields.

gnzlbg (Nov 29 2018 at 09:23, on Zulip):

So I agree with that. Turning those into errors is an orthogonal problems, one has to figure a time line, and none of that is relevant to the RFC.

gnzlbg (Nov 29 2018 at 09:24, on Zulip):

It was a bit unclear to me from the wording, because it appeared that the RFC does make this an error for unsafe code.

gnzlbg (Nov 29 2018 at 09:26, on Zulip):

You can resolve my comment, I don't need the text to be amended with any clarification: your word is enough, and I trust that you don't want to turn the RFC discussion into a discussion about how to phase these errors in.

Last update: Nov 19 2019 at 19:10UTC