regression 1.52 - proc-macro-hack panic

[Mark-Simulacrum]

proc-macro-hack is panicking[1] on beta 1.52 in some cases where it wasn't before:

• base vs. beta
• base vs. beta

My guess is this might have something to do with the usize-suffixed tuple indices, but I'm not sure. It seems to be new behavior on beta; I don't know whether this is caused by any of the recently landed macro-related patches, but it seems not impossible.

[1]: This is my belief, as it's the only repository in the dependency tree to contain the 'source.starts_with(prefix)' string in its source code (based on cargo vendor).

cc @Aaron1011 @petrochenkov @dtolnay

edit - some more cases:

• https://crater-reports.s3.amazonaws.com/beta-1.52-1/beta-2021-03-27/gh/bryanburgers.connectbot/log.txt
• https://crater-reports.s3.amazonaws.com/beta-1.52-1/beta-2021-03-27/gh/guillepb10.performance_test_rust/log.txt
• https://crater-reports.s3.amazonaws.com/beta-1.52-1/beta-2021-03-27/gh/marekm1844.user-tower/log.txt
• https://crater-reports.s3.amazonaws.com/beta-1.52-1/beta-2021-03-27/gh/mindriot101.gitlab-hook-deploy/log.txt
• https://crater-reports.s3.amazonaws.com/beta-1.52-1/beta-2021-03-27/gh/victorinno.tower_web_dummy_service/log.txt
• https://crater-reports.s3.amazonaws.com/beta-1.52-1/beta-2021-03-27/gh/vizitiuRoman.notifier-service/log.txt
• https://crater-reports.s3.amazonaws.com/beta-1.52-1/beta-2021-03-27/gh/vizitiuRoman.rust-notification-service/log.txt
• https://crater-reports.s3.amazonaws.com/beta-1.52-1/beta-2021-03-27/reg/tower-web-protobuf-0.1.0/log.txt
• https://crater-reports.s3.amazonaws.com/beta-1.52-1/beta-2021-03-27/reg/rustiful-0.1.0/log.txt

[Aaron1011]

It looks like the pretty-printing assumptions made by an older version of proc-macro-back have stopped holding. Running cargo update -p proc-macro-hack in one of the failing crates fixed the panic for me.

[Mark-Simulacrum]

This seems pretty widespread (edit: 57 cases, of which 10 are crates.io crates) - I found at least 5-6 more repositories in crater, though not too many actual crates. Is there an easy way for us to continue having the same pretty printing? I know we don't generally guarantee stability there, and it looks like there's a pretty easy upgrade path, so if not that seems OK (but we'll want to mention it in release notes).

[dtolnay]

dtolnay/proc-macro-hack@4df5a94 fixed the whitespace sensitivity for 0.4.3. I have not looked at the build logs but it sounds like the failures are things pinned to an old patch version of 0.4? If so, I would be inclined to close this.

[Mark-Simulacrum]

Yeah, things using 0.4.1 - 57 repositories is not small, and most of these are just tiny self-projects I suspect, so I agree that we can likely close if there's not a trivial fix (probably not, based on the scope of the breakage).

[petrochenkov]

I remember fixing this in the past by tweaking pretty-printing (#63897), not sure what change caused pretty-printing to change again.

[apiraino]

Removing I-prioritize as these panics seems to be solved by upgrading the library (but in case it helps, let's assign a priority).

@rustbot label -I-prioritize

[pietroalbini]

This regression will unfortunately be included in 1.52.0 stable. Updating the labels.

[Mark-Simulacrum]

Closing as "expected breakage"; I don't think we plan to take further action here.