-
Notifications
You must be signed in to change notification settings - Fork 8
Support no_std (Rust 1.64 and later) (Take 2) #14
Conversation
0acaa3c
to
e294368
Compare
Relevant build reports, now that I've made the github actions happy:
(But of course everything else is not yet on 1.64, so several other tests fail). |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I'm suggesting trybuild to support specifying target. And if that proceeds, it would be easier to add and run such tests, and there wouldn't be a need to have an additional check task.
Let's see how that goes.
It's certainly nice to have ready-made actions, but I'm not sure this will save more than the `rustup target add` step -- the dedicated test crate will still be necessary, and only that can be built on the thumb target.
|
No, because |
Short of patching trybuild (which I don't have any capacity to do right now), is there anything I can do to advance this issue? |
Nothing you need to do. I'll probably just merge it when 1.64 reaches stable. |
With 1.64 released, I've been running a round of CI tests at RIOT-OS/rust-riot-wrappers#10 (comment) in parallel to local tests -- unsurprisingly, there were no surprises. I've added one commit updating all mentions of required Rust versions to reflect the strategy indicated in the last comment, and taken the liberty to add a Would you consider making a cstr release soon after merging this? |
New version published. |
Thanks a bunch! |
Closes: #12 (see previous discussion there)
Replaces: #13
Based on the discussion in #13 I've revisited some assumptions, and removed parts of the PR that were based on my wrong understanding of proc-macro crates.
The patches are now in a sequence that, by running CI on them, shows that the test for no-std behavior alone fails, whereas the second patch fixes it.
As before, this breaks support for Rust versions < 1.64, so CI will report all but the "Build on no_std" tests as failing -- so this only makes sense to merge at the point when 1.64 is stable (at which the "nightly" in the test can be switched over).