0.4.1 version bump should be minor version bump about spin-rs HOT 9 CLOSED

The problem is in that Heap is not Send.
We decided to mirror the constraints to the std lib in commit 8d819cc.

As a static variable needs to be Sync, we now require input to the Mutex to be Send before we add a Sync bound ourself. It was a safety bug on our part. See the output of rustc:

src/alloc.rs:64:54: 64:70 note: `*mut u8` cannot be sent between threads safely
src/alloc.rs:64:54: 64:70 note: required because it appears within the type `alloc_buddy_simple::Heap<'static>`
src/alloc.rs:64:54: 64:70 note: required because it appears within the type `core::option::Option<alloc_buddy_simple::Heap<'static>>`
src/alloc.rs:64:54: 64:70 note: shared static variables must have a type that implements `Sync`

While technically correct this breaks semver, since 0.4 is rather new and this is a bug on the others part, I think I am going to leave it this way.

from spin-rs.

Closing due to inactivity.
If you still have some opinion on this, don't hesitate to comment!

from spin-rs.

Sorry, I ran into this just as I went on a trip, but I'm back now.

I understand the rationale for making the change and agree with it.

It's the semver breakage that I'm concerned about (and so are my users). I'm not sure I agree with the "rather new" argument, I developed my code originally against 0.3, but when 0.4.0 came out I noticed everything still worked without any changes in my case so I just updated the dependency to 0.4, expecting it to work in the future.

On the other hand; I'm not sure whether racy code is much better than compile-time errors.

Anyway, no need to dwell on this now, the damage is already done (unless you're willing to make a new 0.4.3 release that's the same as 0.4.0 and continue 0.4.2 as 0.5.0, but that seems cumbersome). I would request you think about breakge when you make changes like these in future though.

from spin-rs.

There is a Cargo blacklist thing that will prevent a specific version from being resolved. (I don't know the name from memory.)

from spin-rs.

I think you mean yank?

from spin-rs.

Yup, that's the one.

from spin-rs.

Thanks for the input.
At this point yanking all 0.4.1+ versions could also cause breakage unfortunately, because of the introduction of new features in 0.4.2.

That 0.4.1 was a breaking change did not occur to me in time. Next time I will be more careful.

from spin-rs.

You could yank 0.4–0.41, which would retroactively make it a proper breaking change. And then the broken ones would be the ones getting the blame too.

from spin-rs.

Not really, because yanking doesn't stop dependencies from being able to download them (no leftpad issues with Rust!)

from spin-rs.