-
Notifications
You must be signed in to change notification settings - Fork 12.7k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Prevent exhaustive matching of Ordering to allow for future extension #37351
Conversation
r? @sfackler (rust_highfive has picked a reviewer for you, use r? to override) |
If the purpose of the PR is to disallow exhaustive matching on |
If we were to do this now I'd prefer to instead just add something like I'd prefer to not embroil this future-proofing with contention over the consume memory ordering, so perhaps we could just do that piece? |
I renamed |
Ok, starting a crater run for this. |
Crater reports shows one regression, but it looks to be suprious. @rfcbot fcp merge |
Team member @alexcrichton has proposed to merge this. The next step is review by the rest of the tagged teams: No concerns currently listed. Once these reviewers reach consensus, this will enter its final comment period. If you spot a major issue that hasn't been raised at any point in this process, please speak up! See this document for info about what commands tagged team members can give me. |
🔔 This is now entering its final comment period, as per the review above. 🔔 psst @alexcrichton, I wasn't able to add the |
@bors: r+ |
📌 Commit 5a2fb88 has been approved by |
Prevent exhaustive matching of Ordering to allow for future extension The C++11 atomic memory model defines a `memory_order_consume` ordering which is generally equivalent to `memory_order_acquire` but can allow better code generation by avoiding memory barrier instructions. Most compilers (including LLVM) currently do not implement this ordering directly and instead treat it identically to `memory_order_acquire`, including adding a memory barrier instruction. There is currently [work](http://open-std.org/Jtc1/sc22/wg21/docs/papers/2016/p0098r1.pdf) to support consume ordering in compilers, and it would be a shame if Rust did not support this. This PR therefore reserves a `__Nonexhaustive` variant in `Ordering` so that adding a new ordering is not a breaking change in the future. This is a [breaking-change] since it disallows exhaustive matching on `Ordering`, however a search of all Rust code on Github shows that there is no code that does this. This makes sense since `Ordering` is typically only used as a parameter to an atomic operation.
The C++11 atomic memory model defines a
memory_order_consume
ordering which is generally equivalent tomemory_order_acquire
but can allow better code generation by avoiding memory barrier instructions. Most compilers (including LLVM) currently do not implement this ordering directly and instead treat it identically tomemory_order_acquire
, including adding a memory barrier instruction.There is currently work to support consume ordering in compilers, and it would be a shame if Rust did not support this. This PR therefore reserves a
__Nonexhaustive
variant inOrdering
so that adding a new ordering is not a breaking change in the future.This is a [breaking-change] since it disallows exhaustive matching on
Ordering
, however a search of all Rust code on Github shows that there is no code that does this. This makes sense sinceOrdering
is typically only used as a parameter to an atomic operation.