1.97 beta regression involving backtraces

[theemathas]

This regression was discovered by the 1.97 beta crater run.

[INFO] [stdout] assertion `left == right` failed
[INFO] [stdout]   left: "╼ inlining[732db11e2f5fb19e]::inlining at /opt/rustwide/workdir/tests/inlining.rs:21:37\n  ├╼ inlining[732db11e2f5fb19e] at /opt/rustwide/workdir/tests/inlining.rs:4:5\n  │  └╼ inlining[732db11e2f5fb19e] at /opt/rustwide/workdir/tests/inlining.rs:10:5\n  └╼ inlining[732db11e2f5fb19e] at /opt/rustwide/workdir/tests/inlining.rs:5:5\n     └╼ inlining[732db11e2f5fb19e] at /opt/rustwide/workdir/tests/inlining.rs:15:5"
[INFO] [stdout]  right: "╼ inlining::inlining::{{closure}} at /opt/rustwide/workdir/tests/inlining.rs:21:37\n  ├╼ inlining::foo at /opt/rustwide/workdir/tests/inlining.rs:4:5\n  │  └╼ inlining::bar at /opt/rustwide/workdir/tests/inlining.rs:10:5\n  └╼ inlining::foo at /opt/rustwide/workdir/tests/inlining.rs:5:5\n     └╼ inlining::baz at /opt/rustwide/workdir/tests/inlining.rs:15:5"

https://crater-reports.s3.amazonaws.com/beta-1.97-3-v2/1.97.0-beta.1/reg/scoped-trace-0.1.0/log.txt

For your convenience, the two strings are formatted properly below:

left (output on beta)

╼ inlining[732db11e2f5fb19e]::inlining at /opt/rustwide/workdir/tests/inlining.rs:21:37
  ├╼ inlining[732db11e2f5fb19e] at /opt/rustwide/workdir/tests/inlining.rs:4:5
  │  └╼ inlining[732db11e2f5fb19e] at /opt/rustwide/workdir/tests/inlining.rs:10:5
  └╼ inlining[732db11e2f5fb19e] at /opt/rustwide/workdir/tests/inlining.rs:5:5
     └╼ inlining[732db11e2f5fb19e] at /opt/rustwide/workdir/tests/inlining.rs:15:5

right (output on stable)

╼ inlining::inlining::{{closure}} at /opt/rustwide/workdir/tests/inlining.rs:21:37
  ├╼ inlining::foo at /opt/rustwide/workdir/tests/inlining.rs:4:5
  │  └╼ inlining::bar at /opt/rustwide/workdir/tests/inlining.rs:10:5
  └╼ inlining::foo at /opt/rustwide/workdir/tests/inlining.rs:5:5
     └╼ inlining::baz at /opt/rustwide/workdir/tests/inlining.rs:15:5

Unfortunately, I am having trouble reproducing this locally, which I suspect that might be caused by the scoped-trace crate requiring a certain setup to function properly. (Locally, on my computer, the left string is an empty string on both stable and beta.)

[theemathas]

Another regression with probably the same root cause:

[INFO] [stdout] thread 'stacktrace_util::tests::stacktrace_from_method' (17) panicked at src/stacktrace_util.rs:168:9:
[INFO] [stdout] method name: rust_debugging_locks[f42667667beee1e2]::stacktrace_util::tests::caller_function

https://crater-reports.s3.amazonaws.com/beta-1.97-3-v2/1.97.0-beta.1/reg/rust-debugging-locks-0.8.9/log.txt

The code expected the string rust_debugging_locks[f42667667beee1e2]::stacktrace_util::tests::caller_function to start with rust_debugging_locks::stacktrace_util::tests::caller_function::h

The regression does reproduce properly on my local setup though.

[bjorn3]

Both are probably just fallout from #151994. The v0 mangling schemes use a different symbol hash location. The regression in OP seems like it is caused by removing the last component to hide the symbol hash, but the symbol hash is no longer there, so it truncates the actual function name.

[theemathas]

Yet another crate comparing backtrace strings: https://crater-reports.s3.amazonaws.com/beta-1.97-3-v2/1.97.0-beta.1/gh/yisun-git.Hypervisor_on_FC/log.txt

Relevant code: https://github.com/yisun-git/Hypervisor_on_FC/blob/6a6d8d11b360f4984e1700965ef735731a59fca1/src/main.rs#L278

[theemathas]

https://crater-reports.s3.amazonaws.com/beta-1.97-3-v2/1.97.0-beta.1/gh/harwinder349.Github/log.txt Appears to be the same code as above comment.

[theemathas]

I've filed issues on each affected repository:

• Incoming breakage in Rust 1.97.0 jswrenn/scoped-trace#3
• Incoming breakage in Rust 1.97.0 grooviegermanikus/rust-debugging-locks#1
• Incoming breakage in Rust 1.97.0 yisun-git/Hypervisor_on_FC#1
• Incoming breakage in Rust 1.97.0 harwinder349/Github#7

Should this issue be closed as accepted breakage?