Stream: t-compiler/const-eval

Topic: #55549 value visitors


RalfJ (Nov 06 2018 at 12:28, on Zulip):

@Oli I don't get your method proposal

RalfJ (Nov 06 2018 at 12:28, on Zulip):

are you saying it should be a method returning bool?

RalfJ (Nov 06 2018 at 12:29, on Zulip):

If it is a method to call on the list fields before iteration, I think I'd rather remove this entirely and do sorting on the miri side.

RalfJ (Nov 06 2018 at 12:29, on Zulip):

that's probably more elegant anyway...

RalfJ (Nov 06 2018 at 12:31, on Zulip):

hm, except that it would be even nicer to use https://doc.rust-lang.org/nightly/nightly-rustc/rustc_target/abi/enum.FieldPlacement.html#method.index_by_increasing_offset

oli (Nov 06 2018 at 12:34, on Zulip):

I do dislike having the code in the miri-engine. I meant a method that does the sorting, yes.

oli (Nov 06 2018 at 12:35, on Zulip):

I mean you need a Vec at the end anyway, right? So there's no problem with impl trait or sth

oli (Nov 06 2018 at 12:35, on Zulip):

the code in miri can use index_by_increasing_offset, and miri-engine can just do the regular thing

oli (Nov 06 2018 at 12:35, on Zulip):

or

oli (Nov 06 2018 at 12:36, on Zulip):

do we expect the miri-engine to sort some day in the future, too?

oli (Nov 06 2018 at 12:36, on Zulip):

if so, just unconditionally sort with the comment as an explanation

RalfJ (Nov 06 2018 at 13:49, on Zulip):

I'll just revert that commit and do it all in miri without a new hook, seems simpler

RalfJ (Nov 06 2018 at 13:49, on Zulip):

abstracting over index_by_increasing_offset does not worki because I cannot use impl Trait in a trait

Last update: Nov 15 2019 at 20:30UTC