Avoid panic branch in EscapeDefault::backslash and EscapeDebug::backslash.

[reitermarkus]

See japaric/ufmt#52 and https://github.com/andrewgazelka/ufmt/blob/e97ce1a86cfcc9fa36ffab6ee62762c00e4b3fc9/src/impls/core.rs#L62.

This avoids the unreachable panicking branch in char::escape_debug.

Not sure what the best way to test this is.

[rustbot]

r? @m-ou-se

rustbot has assigned @m-ou-se.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

[reitermarkus]

r?

[rustbot]

Error: Parsing assign command in comment failed: ...'' | error: specify user to assign to at >| ''...

Please file an issue on GitHub at triagebot if there's a problem with this bot, or reach out on #t-infra on Zulip.

[joboet]

Are you sure that that's where the panic comes from? As far as I can tell, LLVM is able to optimize this out.

[reitermarkus]

I assume the panic_never crate (used to test ufmt) only works before optimization, i.e. it won't compile (or more specifically link, I guess) if there is any panicking branch, even if it is eventually optimized out.

[joboet]

This is the wrong panic, the assert that cannot be optimized out is in ExactSizeIterator::len.

See how the BB in bjorn's Godbolt example is only reached from two comparisons that depend on the result of EscapeIterInner::size_hint? I think this is this assertion:

rust/library/core/src/iter/traits/exact_size.rs

Line 122 in 9db7a74

assert_eq!(upper, Some(lower));

It will never trigger but cannot be optimized out since EscapeIterInner::size_hint isn't marked #[inline]:

rust/library/core/src/char/mod.rs

Line 342 in 9db7a74

fn size_hint(&self) -> (usize, Option<usize>) {

I'll close this PR, but feel free to r? me on a PR that adds the #[inline] hint.

[reitermarkus]

Thanks for looking into this, @joboet, I have opened a PR to add the inline attribute here: #124307