Add #[allow_internal_unstable] to track stability for macros better. by huonw · Pull Request #22899 · rust-lang/rust

huonw · 2015-02-28T14:42:09Z

Unstable items used in a macro expansion will now always trigger
stability warnings, unless the unstable items are directly inside a
macro marked with #[allow_internal_unstable]. IOW, the compiler warns
unless the span of the unstable item is a subspan of the definition of a
macro marked with that attribute.

E.g.

#[allow_internal_unstable]
macro_rules! foo {
($e: expr) => {{
$e;
unstable(); // no warning
only_called_by_foo!();
}}
}
macro_rules! only_called_by_foo {
() => { unstable() } // warning
}
foo!(unstable()) // warning

The unstable inside foo is fine, due to the attribute. But the
unstable inside only_called_by_foo is not, since that macro doesn't
have the attribute, and the unstable passed into foo is also not
fine since it isn't contained in the macro itself (that is, even though
it is only used directly in the macro).

In the process this makes the stability tracking much more precise,
e.g. previously println!("{}", unstable()) got no warning, but now it
does. As such, this is a bug fix that may cause [breaking-change]s.

The attribute is definitely feature gated, since it explicitly allows
side-stepping the feature gating system.

This updates thread_local! macro to use the attribute, since it uses
unstable features internally (initialising a struct with unstable
fields).

rust-highfive · 2015-02-28T14:42:12Z

r? @kmcallister

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

huonw · 2015-02-28T14:42:56Z

r? @alexcrichton, @aturon

@alexcrichton this is the fix for the macro issues we were discussing on IRC a few days ago.

aturon · 2015-02-28T18:03:29Z

🏆

alexcrichton · 2015-02-28T18:41:54Z

Nice! This looks pretty great to me. Can this de-stabilize all of the internal fields of the thread local structs as well? (now that they can explicitly be left as unstable)

huonw · 2015-03-01T03:15:21Z

Updated, destabilising essentially everything in thread_local other than the main macro and the Key type.

alexcrichton · 2015-03-01T06:58:16Z

@bors: r+ 5ea24dd

Manishearth · 2015-03-01T08:17:22Z

This is awesome :D

bors · 2015-03-01T21:01:09Z

⌛ Testing commit 5ea24dd with merge 2196aa3...

bors · 2015-03-01T22:02:51Z

💔 Test failed - auto-mac-32-opt

Manishearth · 2015-03-01T22:04:15Z

run-pass/macro-attributes.rs needs a #![feature(custom_attribute)]

huonw · 2015-03-02T03:38:01Z

@bors r=alexcrichton 49ef

bors · 2015-03-02T10:07:51Z

⌛ Testing commit 49ef199 with merge 29c07e2...

bors · 2015-03-02T11:08:44Z

💔 Test failed - auto-mac-32-opt

Manishearth · 2015-03-02T11:10:51Z


---- [run-pass] run-pass-fulldeps/plugin-args-1.rs stdout ----
error: auxiliary build of "/Users/rustbuild/src/rust-buildbot/slave/auto-mac-32-opt/build/src/test/auxiliary/plugin_args.rs" failed to compile: status: exit code: 101
command: i686-apple-darwin/stage2/bin/rustc /Users/rustbuild/src/rust-buildbot/slave/auto-mac-32-opt/build/src/test/auxiliary/plugin_args.rs -L i686-apple-darwin/test/run-pass-fulldeps --target=i686-apple-darwin --crate-type=dylib -L i686-apple-darwin/test/run-pass-fulldeps/plugin-args-1.stage2-i686-apple-darwinlibaux -C prefer-dynamic --out-dir i686-apple-darwin/test/run-pass-fulldeps/plugin-args-1.stage2-i686-apple-darwinlibaux --cfg rtopt --cfg debug -O -L i686-apple-darwin/rt
stdout:
------------------------------------------
------------------------------------------
stderr:
------------------------------------------
/Users/rustbuild/src/rust-buildbot/slave/auto-mac-32-opt/build/src/test/auxiliary/plugin_args.rs:49:9: 49:53 error: this function takes 3 parameters but 2 parameters were supplied [E0061]
/Users/rustbuild/src/rust-buildbot/slave/auto-mac-32-opt/build/src/test/auxiliary/plugin_args.rs:49 NormalTT(box Expander { args: args, }, None));
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
error: aborting due to previous error
------------------------------------------
thread '[run-pass] run-pass-fulldeps/plugin-args-1.rs' panicked at 'explicit panic', /Users/rustbuild/src/rust-buildbot/slave/auto-mac-32-opt/build/src/compiletest/runtest.rs:1466
---- [run-pass] run-pass-fulldeps/plugin-args-2.rs stdout ----
error: auxiliary build of "/Users/rustbuild/src/rust-buildbot/slave/auto-mac-32-opt/build/src/test/auxiliary/plugin_args.rs" failed to compile: status: exit code: 101
command: i686-apple-darwin/stage2/bin/rustc /Users/rustbuild/src/rust-buildbot/slave/auto-mac-32-opt/build/src/test/auxiliary/plugin_args.rs -L i686-apple-darwin/test/run-pass-fulldeps --target=i686-apple-darwin --crate-type=dylib -L i686-apple-darwin/test/run-pass-fulldeps/plugin-args-2.stage2-i686-apple-darwinlibaux -C prefer-dynamic --out-dir i686-apple-darwin/test/run-pass-fulldeps/plugin-args-2.stage2-i686-apple-darwinlibaux --cfg rtopt --cfg debug -O -L i686-apple-darwin/rt
stdout:
------------------------------------------
------------------------------------------
stderr:
------------------------------------------
/Users/rustbuild/src/rust-buildbot/slave/auto-mac-32-opt/build/src/test/auxiliary/plugin_args.rs:49:9: 49:53 error: this function takes 3 parameters but 2 parameters were supplied [E0061]
/Users/rustbuild/src/rust-buildbot/slave/auto-mac-32-opt/build/src/test/auxiliary/plugin_args.rs:49 NormalTT(box Expander { args: args, }, None));
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
error: aborting due to previous error
------------------------------------------
thread '[run-pass] run-pass-fulldeps/plugin-args-2.rs' panicked at 'explicit panic', /Users/rustbuild/src/rust-buildbot/slave/auto-mac-32-opt/build/src/compiletest/runtest.rs:1466
---- [run-pass] run-pass-fulldeps/plugin-args-3.rs stdout ----
error: auxiliary build of "/Users/rustbuild/src/rust-buildbot/slave/auto-mac-32-opt/build/src/test/auxiliary/plugin_args.rs" failed to compile: status: exit code: 101
command: i686-apple-darwin/stage2/bin/rustc /Users/rustbuild/src/rust-buildbot/slave/auto-mac-32-opt/build/src/test/auxiliary/plugin_args.rs -L i686-apple-darwin/test/run-pass-fulldeps --target=i686-apple-darwin --crate-type=dylib -L i686-apple-darwin/test/run-pass-fulldeps/plugin-args-3.stage2-i686-apple-darwinlibaux -C prefer-dynamic --out-dir i686-apple-darwin/test/run-pass-fulldeps/plugin-args-3.stage2-i686-apple-darwinlibaux --cfg rtopt --cfg debug -O -L i686-apple-darwin/rt
stdout:
------------------------------------------
------------------------------------------
stderr:
------------------------------------------
/Users/rustbuild/src/rust-buildbot/slave/auto-mac-32-opt/build/src/test/auxiliary/plugin_args.rs:49:9: 49:53 error: this function takes 3 parameters but 2 parameters were supplied [E0061]
/Users/rustbuild/src/rust-buildbot/slave/auto-mac-32-opt/build/src/test/auxiliary/plugin_args.rs:49 NormalTT(box Expander { args: args, }, None));
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
error: aborting due to previous error
------------------------------------------
thread '[run-pass] run-pass-fulldeps/plugin-args-3.rs' panicked at 'explicit panic', /Users/rustbuild/src/rust-buildbot/slave/auto-mac-32-opt/build/src/compiletest/runtest.rs:1466
failures:
[run-pass] run-pass-fulldeps/plugin-args-1.rs
[run-pass] run-pass-fulldeps/plugin-args-2.rs
[run-pass] run-pass-fulldeps/plugin-args-3.rs

bors · 2015-03-03T14:21:17Z

☔ The latest upstream changes (presumably #22995) made this pull request unmergeable. Please resolve the merge conflicts.

huonw · 2015-03-04T07:23:12Z

Updated; I've modified the approach and several extra commits (the fixup! ... one are intended as a corrections to the earlier commit, preserved for to allow you to see what changed. I'll rebase and squash post review).

I've had to add another feature gating pass, because the post-expansion feature-gating pass ran before the second configuration pass, and so would miss custom attributes inserted by cfg_attr's inside macro expansion: huonw@e397c4a. (I'm sure this can be done in a more rigorous way, but, it seems sensible to run feature gating on exactly the AST we compile, which this extra pass does.)

alexcrichton · 2015-03-04T18:18:18Z

This looks good to me, r=me.

Could you also knock out some #[stable] attributes in sys_common::thread_local as well as part of this PR? They're reexported through std::thread_local::__imp currently.

theemathas · 2015-03-05T04:58:11Z

What about select! {}?

huonw · 2015-03-05T11:56:20Z

select! is unstable.

Unstable items used in a macro expansion will now always trigger stability warnings, *unless* the unstable items are directly inside a macro marked with `#[allow_internal_unstable]`. IOW, the compiler warns unless the span of the unstable item is a subspan of the definition of a macro marked with that attribute. E.g. #[allow_internal_unstable] macro_rules! foo { ($e: expr) => {{ $e; unstable(); // no warning only_called_by_foo!(); }} } macro_rules! only_called_by_foo { () => { unstable() } // warning } foo!(unstable()) // warning The unstable inside `foo` is fine, due to the attribute. But the `unstable` inside `only_called_by_foo` is not, since that macro doesn't have the attribute, and the `unstable` passed into `foo` is also not fine since it isn't contained in the macro itself (that is, even though it is only used directly in the macro). In the process this makes the stability tracking much more precise, e.g. previously `println!("{}", unstable())` got no warning, but now it does. As such, this is a bug fix that may cause [breaking-change]s. The attribute is definitely feature gated, since it explicitly allows side-stepping the feature gating system.

This destabilises all the implementation details of `thread_local!`, since they do not *need* to be stable with the new attribute.

This ensures we catch everything; previously, an unknown attribute inserted by #[cfg_attr(...)] in a macro expansion would not be detected.

huonw · 2015-03-05T13:28:45Z

@bors r=alexcrichton b5c6

bors · 2015-03-05T14:25:51Z

⌛ Testing commit b5c6ab2 with merge ffb90ea...

bors · 2015-03-05T14:31:26Z

💔 Test failed - auto-linux-64-x-android-t

alexcrichton · 2015-03-05T17:35:21Z

@bors: retry

alexcrichton · 2015-03-05T17:54:52Z

@bors: retry

Unstable items used in a macro expansion will now always trigger stability warnings, *unless* the unstable items are directly inside a macro marked with `#[allow_internal_unstable]`. IOW, the compiler warns unless the span of the unstable item is a subspan of the definition of a macro marked with that attribute. E.g. #[allow_internal_unstable] macro_rules! foo { ($e: expr) => {{ $e; unstable(); // no warning only_called_by_foo!(); }} } macro_rules! only_called_by_foo { () => { unstable() } // warning } foo!(unstable()) // warning The unstable inside `foo` is fine, due to the attribute. But the `unstable` inside `only_called_by_foo` is not, since that macro doesn't have the attribute, and the `unstable` passed into `foo` is also not fine since it isn't contained in the macro itself (that is, even though it is only used directly in the macro). In the process this makes the stability tracking much more precise, e.g. previously `println!(\"{}\", unstable())` got no warning, but now it does. As such, this is a bug fix that may cause [breaking-change]s. The attribute is definitely feature gated, since it explicitly allows side-stepping the feature gating system. --- This updates `thread_local!` macro to use the attribute, since it uses unstable features internally (initialising a struct with unstable fields).

bors · 2015-03-06T05:20:12Z

⌛ Testing commit b5c6ab2 with merge 1fe8f22...

Unstable items used in a macro expansion will now always trigger stability warnings, *unless* the unstable items are directly inside a macro marked with `#[allow_internal_unstable]`. IOW, the compiler warns unless the span of the unstable item is a subspan of the definition of a macro marked with that attribute. E.g. #[allow_internal_unstable] macro_rules! foo { ($e: expr) => {{ $e; unstable(); // no warning only_called_by_foo!(); }} } macro_rules! only_called_by_foo { () => { unstable() } // warning } foo!(unstable()) // warning The unstable inside `foo` is fine, due to the attribute. But the `unstable` inside `only_called_by_foo` is not, since that macro doesn't have the attribute, and the `unstable` passed into `foo` is also not fine since it isn't contained in the macro itself (that is, even though it is only used directly in the macro). In the process this makes the stability tracking much more precise, e.g. previously `println!("{}", unstable())` got no warning, but now it does. As such, this is a bug fix that may cause [breaking-change]s. The attribute is definitely feature gated, since it explicitly allows side-stepping the feature gating system. --- This updates `thread_local!` macro to use the attribute, since it uses unstable features internally (initialising a struct with unstable fields).

bors · 2015-03-06T07:59:20Z

☀️ Test successful - auto-linux-32-nopt-t, auto-linux-32-opt, auto-linux-64-nopt-t, auto-linux-64-opt, auto-linux-64-x-android-t, auto-mac-32-opt, auto-mac-64-nopt-t, auto-mac-64-opt, auto-win-32-nopt-t, auto-win-32-opt, auto-win-64-nopt-t, auto-win-64-opt

fix: Support macros in `#[doc]` attributes in IDE features

rust-highfive assigned alexcrichton Feb 28, 2015

huonw force-pushed the macro-stability branch from 903959c to 5ea24ddCompare March 1, 2015 03:14

huonw force-pushed the macro-stability branch from 5ea24dd to 49ef199Compare March 2, 2015 03:37

huonw mentioned this pull request Mar 2, 2015
Destabilise the internals of format_args!#22953
Closed

huonw force-pushed the macro-stability branch from 49ef199 to 32bc3d5Compare March 4, 2015 07:11

huonw force-pushed the macro-stability branch from 63cedcd to e397c4aCompare March 4, 2015 11:35

huonw force-pushed the macro-stability branch from e397c4a to 89a2709Compare March 5, 2015 12:04

huonw added 3 commits March 6, 2015 00:18

Use #[allow_internal_unstable] for thread_local!
ab7ef74
This destabilises all the implementation details of `thread_local!`, since they do not *need* to be stable with the new attribute.

Add more debugging to syntax::feature_gate.
10426f6

Run feature-gating on the final AST passed to the compiler.
b5c6ab2
This ensures we catch everything; previously, an unknown attribute inserted by #[cfg_attr(...)] in a macro expansion would not be detected.

huonw force-pushed the macro-stability branch from 89a2709 to b5c6ab2Compare March 5, 2015 13:18

alexcrichton mentioned this pull request Mar 5, 2015
RFC: Deprecate std::fmt::format in favor of String::format rust-lang/rfcs#810
Closed

bors merged commit b5c6ab2 into rust-lang:masterMar 6, 2015

bors mentioned this pull request Mar 6, 2015
overloaded-box protocol changes #22086
Closed

sfackler mentioned this pull request Mar 9, 2015
Make #[derive(Anything)] into sugar for #[derive_Anything] #23137
Merged

huonw deleted the macro-stability branch March 10, 2015 04:37

jseyfried mentioned this pull request Apr 8, 2016
Macro-expanded unconfigured items are gated feature checked #32840
Closed

lnicola pushed a commit to lnicola/rust that referenced this pull request Aug 10, 2026
Merge pull requestrust-lang#22899from ChayimFriedman2/doc-macro-ide
75d2068
fix: Support macros in `#[doc]` attributes in IDE features

Uh oh!

Uh oh!

Conversation

huonw commented Feb 28, 2015

Uh oh!

rust-highfive commented Feb 28, 2015

Uh oh!

huonw commented Feb 28, 2015

Uh oh!

aturon commented Feb 28, 2015

Uh oh!

alexcrichton commented Feb 28, 2015

Uh oh!

huonw commented Mar 1, 2015

Uh oh!

alexcrichton commented Mar 1, 2015

Uh oh!

Manishearth commented Mar 1, 2015

Uh oh!

bors commented Mar 1, 2015

Uh oh!

bors commented Mar 1, 2015

Uh oh!

Manishearth commented Mar 1, 2015

Uh oh!

huonw commented Mar 2, 2015

Uh oh!

bors commented Mar 2, 2015

Uh oh!

bors commented Mar 2, 2015

Uh oh!

Manishearth commented Mar 2, 2015

Uh oh!

bors commented Mar 3, 2015

Uh oh!

huonw commented Mar 4, 2015

Uh oh!

alexcrichton commented Mar 4, 2015

Uh oh!

theemathas commented Mar 5, 2015

Uh oh!

huonw commented Mar 5, 2015

Uh oh!

huonw commented Mar 5, 2015

Uh oh!

bors commented Mar 5, 2015

Uh oh!

bors commented Mar 5, 2015

Uh oh!

alexcrichton commented Mar 5, 2015

Uh oh!

alexcrichton commented Mar 5, 2015

Uh oh!

bors commented Mar 6, 2015

Uh oh!

bors commented Mar 6, 2015

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

8 participants