replace `convert::Infallible` with `!`

[canndrew]

No description provided.

[rust-highfive]

r? @sfackler

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

[scottmcm]

Horray! One step closer to stable TryFrom...

[scottmcm]

Horray! One step closer to stable TryFrom...

[scottmcm]

Maybe leave this as impl From<!> For TryFromIntError? This is here so that you can use u32: TryFrom<u16> in a method returning Result<_, TryFromIntError>, which I think is still wanted. (Not for those types literally, but for when they're c_whatever aliases.)

[canndrew]

I removed it thinking that we'd have impl<T> From<!> for T, but since we won't have that until we have specialization with intersection impls I guess I'd better add it back.

[scottmcm]

Looking through that list of derives, are we missing Copy and Clone for !? They're not in the rustdoc, at least: https://doc.rust-lang.org/nightly/std/primitive.never.html#implementations

(Not a blocker for this PR, of course.)

[canndrew]

They're implemented. No idea why they're not showing up in rustdoc.

[scottmcm]

Ah, looks like this is a nearing-three-year-old bug with primitives: #25893

[canndrew]

It looks like the travis failure here had nothing to do with this PR.

[SimonSapin]

tidy error: /checkout/src/libcore/num/mod.rs:3598: trailing whitespace

[canndrew]

Oops, thanks for pointing that out.

[SimonSapin]

[01:27:14] ---- [rustdoc] rustdoc/synthetic_auto/no-redundancy.rs stdout ----
[01:27:14] 	
[01:27:14] error: htmldocck failed!
[01:27:14] status: exit code: 1
[01:27:14] command: "/usr/bin/python2.7" "/checkout/src/etc/htmldocck.py" "/checkout/obj/build/x86_64-unknown-linux-gnu/test/rustdoc/synthetic_auto/no-redundancy.stage2-x86_64-unknown-linux-gnu" "/checkout/src/test/rustdoc/synthetic_auto/no-redundancy.rs"
[01:27:14] stdout:
[01:27:14] ------------------------------------------
[01:27:14] 
[01:27:14] ------------------------------------------
[01:27:14] stderr:
[01:27:14] ------------------------------------------
[01:27:14] 22: @has check failed
[01:27:14] thread 'main' panicked at 'Some tests failed', tools/compiletest/src/main.rs:478:22
[01:27:14] 	`XPATH PATTERN` did not match
[01:27:14] 	// @has - '//*[@id="synthetic-implementations-list"]/*[@class="impl"]/*/code' "impl<T> Send for Outer<T> where T: Copy + Send"
[01:27:14] 
[01:27:14] Encountered 1 errors
[01:27:14] 
[01:27:14] ------------------------------------------
[01:27:14] 
[01:27:14] thread '[rustdoc] rustdoc/synthetic_auto/no-redundancy.rs' panicked at 'explicit panic', tools/compiletest/src/runtest.rs:2903:9
[01:27:14] 
[01:27:14] 
[01:27:14] failures:
[01:27:14]     [rustdoc] rustdoc/synthetic_auto/no-redundancy.rs

This looks like the same as #48265 (comment), you may need to swap the order of bounds in the expected rustdoc output in the comment in src/test/rustdoc/synthetic_auto/no-redundancy.rs.

Something like ./x.py test src/test/rustdoc --test-args no-redundancy should run just this test.

[SimonSapin]

I’ve filed #49123 for that rustdoc test failure.

[SimonSapin]

#49058 disabled that failing test.

@bors: r+

[bors]

📌 Commit 15bab45 has been approved by SimonSapin

[F001]

There are several empty enum types in std. For example:

impl FromStr for String {
  type Err = ParseError;
}

My question is, are they going to be replaced by ! type ?
The type pub enum ParseError {} has been stabilized. Is it a breaking change if we replace it by ! ?

[SimonSapin]

This is being discussed at #49039