Replace all fmt.pad with debug_struct

[CDirkx]

This replaces any occurrence of:

• f.pad("X") with f.debug_struct("X").finish()
• f.pad("X { .. }") with f.debug_struct("X").finish_non_exhaustive()

This is in line with existing formatting code such as

rust/library/std/src/sync/mpsc/mod.rs

Lines 1470 to 1475 in 1255053

	#[stable(feature = "mpsc_debug", since = "1.8.0")]
	impl<T> fmt::DebugforReceiver<T>{
	fnfmt(&self,f:&mut fmt::Formatter<'_>) -> fmt::Result{
	f.debug_struct("Receiver").finish_non_exhaustive()
	}
	}

[rust-highfive]

r? @kennytm

(rust-highfive has picked a reviewer for you, use r? to override)

[kennytm]

i guess this is fine since Debug output is not guaranteed to be stable, but note that pad and debug_struct produce different output when padding is involved:

use std::fmt;structA;structB;impl fmt::DebugforA{fnfmt(&self,f:&mut fmt::Formatter<'_>) -> fmt::Result{
f.debug_struct("X").finish()}}impl fmt::DebugforB{fnfmt(&self,f:&mut fmt::Formatter<'_>) -> fmt::Result{
f.pad("X")}}fnmain(){assert_eq!(format!("<{:3?}>",A),"<X>");assert_eq!(format!("<{:3?}>",B),"<X >");}

(sending this to a real T-libs member) r? @m-ou-se

[m-ou-se]

Some of these should probably use finish_non_exhaustive instead of finish. E.g. UnsafeCell { .. } instead of UnsafeCell since it contains stuff. Same for Any I suppose.

(I replaced a bunch of those recently in #83558, but I missed these since they used .pad() instead of .debug_struct().)

[m-ou-se]

i guess this is fine since Debug output is not guaranteed to be stable, but note that pad and debug_struct produce different output when padding is involved:

We somewhat regularly change or improve Debug output, so I'm not worried about stability here. Consistency with all other Debug implementations is useful. The other Debug implementations (including those from #[derive(Debug)]) forward the formatting options to each field individually, and don't apply it to the type names:

#[derive(Debug)]structA{x:i32,y:i32,}fnmain(){println!("{:#010?}",A{ x:1, y:2});}

A {
x: 0000000001,
y: 0000000002,
}

The only downside of the change in this PR is that in 'pretty' mode ({:#?}, as used by dbg!()), .finish_non_exhaustive puts the .. on a separate line:

A {
..
}

taking up a lot more space than A { .. } like before for these types. But that's something we can fix in the .finish_non_exhaustive() implementation itself: #84390

[CDirkx]

I agree with the use of finish_non_exhaustive for UnsafeCell and Any. I checked the others, but those two are the only ones with meaningful internal data that should be represented (so excluding PhantomData).

I also found two more cases (PoisonError and SendError) that were using "X { .. }".fmt(f).

[m-ou-se]

Thanks!

@bors r+

[bors]

📌 Commit fdae757 has been approved by m-ou-se

[m-ou-se]

@bors r-

[CDirkx]

I fixed alloc::test::test_show for now, by changing assert_eq!(s, "Any") to assert_eq!(s, "Any { .. }"). It's still fragile to any future changes to the AnyDebug implementations, but reworking the test could be done in a future PR.

[m-ou-se]

@bors r+

[bors]

📌 Commit fccc75c has been approved by m-ou-se

[bors]

🌲 The tree is currently closed for pull requests below priority 1000. This pull request will be tested once the tree is reopened.