Updated RELEASES.md for 1.21.0

[XAMPPRocky]

Rendered

[rust-highfive]

r? @alexcrichton

(rust_highfive has picked a reviewer for you, use r? to override)

[petrochenkov]

typo [T; N]

[petrochenkov]

Foo::<u8>(0) was always possible, what #43540 does is permitting (sometimes unnecessary) :: before < in all paths, including paths passed to macros.
Previously:

my_macro!(Vec<i32>::new); // OK
my_macro!(Vec::<i32>::new); // ERROR

Now:

my_macro!(Vec<i32>::new); // OK
my_macro!(Vec::<i32>::new); // OK too

[zackmdavis]

Release note readers shouldn't need to think about named internal implementation details like CodeMap, and I'm not sure where "which required unstable code" is coming from.

Maybe something like "was previously relative to a concatenation of files in a crate"?? (I don't actually know how the compiler determines which files go in a CodeMap—and neither does typical consumer of error-message JSON, which is why we fixed this).

[zackmdavis]

There should also be a compatibility bullet point for the unused_results lint getting stronger. "unused_results lint no longer ignores booleans"?

[kennytm]

Is there any breakage examples?

BTW "trait selection impl in method::probe" sounds too technical for readers.

[XAMPPRocky]

It definitely is I just couldn't think of anything better. @arielb1 would you know a better explanation and example?

[kennytm]

@Aaronepower: Maybe summarize or just link to #44224 (comment) ?

[XAMPPRocky]

@zackmdavis Could you provide a PR link? I can't seem to find the PR that makes the lint stronger.

[alexcrichton]

Thanks @Aaronepower!

I think this may also want to mention [patch] in Cargo (detailed in RFC 1969)

[alexcrichton]

Thanks @Aaronepower!

I think this may also want to mention [patch] in Cargo (detailed in RFC 1969)

[alexcrichton]

We may not want to highlight this just yet as they're still a redirect to doc.crates.io, but soon though!

[XAMPPRocky]

I didn't think so either, it was marked as relnotes so I included it.

[alexcrichton]

Another Cargo change that may be worth mentioning is rust-lang/cargo#4270, a change to the ignore/exclude syntax

[alexcrichton]

Also rust-lang/cargo#4400, the addition of the --all-targets flag

[alexcrichton]

Also rust-lang/cargo#4364, a new compatibility note

[alexcrichton]

Speaking of lint changes, I believe unused_extern_crate was turned on by default in #42588

[zackmdavis]

@Aaronepower

Could you provide a PR link? I can't seem to find the PR that makes the lint stronger.

The change was, somewhat embarrassingly, unintentional in #43728 (removed a TyBool => return match arm without realizing that this would let us "fall through" to the part of the lint-pass that sets unused_results), but then the discussion and final-comment-period in #44119 retroactively ruled it a bugfix rather than a regression.

[ollie27]

Looks like #43838 is missing.

[ollie27]

Looks like #43838 is missing.

[ollie27]

It's more than this. Clone is now implemented for all arrays with T: Clone, all tuples with every member being Clone and all function pointers.

[ollie27]

missing link

[alexcrichton]

gah sorry for letting this sit @Aaronepower! slipped off my radar :(

In any case looks great, thanks so much again for tackling this! Let's get this in and we can continue to iterate in-tree

@bors: r+

[bors]

📌 Commit f3c316d has been approved by alexcrichton

[bors]

⌛️ Testing commit f3c316d with merge 547d8a3...

[bors]

💔 Test failed - status-appveyor

[kennytm]

@bors retry Wat.

Spurious network error, can't resolve crates.io.

[bors]

⌛️ Testing commit f3c316d with merge 65e8d5a...

[nrc]

Adds the RLS! rustup component add rls-preview

[XAMPPRocky]

@nrc I've added that, let know if the PR I picked is the wrong one to link.

[kennytm]

This should be &0, not 0.

[dtolnay]

We are stabilizing std::mem::discriminant in 1.21. #44263, #44552

We could backport any of the following library stabilizations -- I'm not clear how we decide whether to do that.

• TcpStream::connect_timeout #44563
• Iterator::for_each #44567
• Ord::min/max #44593
• std::sync::atomic::compiler_fence #44595
• std::mem::needs_drop #44639
• Vec::splice #44640

[QuietMisdreavus]

The docs got a few notable quality-of-life additions. Not sure whether they're all "release notes" worthy, but i'll note them here all the same:

• std::os docs for Unix, Linux, and Windows now appear in all docs, including the ones online
• references and function pointers have their own primitive pages in the std docs
• trait implementors now print their associated types without having to go to the type's page to find them

[michaelwoerister]

I think the "async-llvm" changes from #43506 will be in the next release (right, @alexcrichton?). This feature is pretty neat and it solves a fundamental problem we started running into with incremental compilation: Before, the amount of memory consumed by the LLVM part of compilation was determined by the total size of your crate. After, it is determined by the size of the largest N codegen-units, where N is the number of parallel threads you tell rustc to use. Note that this is something especially important for incr. comp. but the feature helps with any setup that uses codegen-units=N.

Together with @alexcrichton's jobserver changes, this makes using something like codegen-units= 2 * cpu-core-count much more viable, if runtime performance is not the highest priority (e.g. most kinds of development builds).

[arielb1]

This is being backed out (#44825)

[alexcrichton]

Alright the release is just around the corner, so let's merge!

@bors: r+

[bors]

📌 Commit 7ed23f1 has been approved by alexcrichton

[bors]

⌛️ Testing commit 7ed23f1 with merge bd36dcf...

[bors]

☀️ Test successful - status-appveyor, status-travis
Approved by: alexcrichton
Pushing bd36dcf to master...