Backtrace and Spantrace support about miette

[zkat]

miette does have backtrace printing/hook capabilities for panics. What I'm looking for is support for printing backtraces from backtrace fields in errors. From thiserror docs:

anyhow then looks for these provided backtraces and prints them. From the anyhow docs:

[erratic-pattern]

@zkat I've been working on some basic backtrace stuff with miette as a library user, but I could try to adapt it into a proper feature at some point.

backtrace code I pulled directly from human-panic, which is also what eyre uses last I checked: https://github.com/cloudtruth/cloudtruth-cli/blob/8ba6b5a4bfde7684cbeac128f68fa4f4f4366408/crates/integration-test-harness/src/backtrace.rs#L13

And then I created a very basic Diagnostic for panic messages with an optional backtrace: https://github.com/cloudtruth/cloudtruth-cli/blob/8ba6b5a4bfde7684cbeac128f68fa4f4f4366408/crates/integration-test-harness/src/panic.rs#L55

What do you think the best approach is for adding something like this to the library? Something similar to eyre does it where it just checks for RUST_BACKTRACE and RUST_SPANTRACE? Also a fancier backtrace renderer is probably in order.

[erratic-pattern]

The use case with backtraces that I built on top of miette is to actually show source code from failing unit tests. To do this I use track_caller to find the source file of the code that installed the panic handler, and then search the backtrace for the first frame matching that directory path. See https://github.com/cloudtruth/cloudtruth-cli/blob/8ba6b5a4bfde7684cbeac128f68fa4f4f4366408/crates/integration-test-harness/src/source_span.rs#L75

The install function is called on each test from a #[use_test_harness] attribute that I wrote.

Output looks like this:

Maybe this could be a second crate that someone could choose to pull in if they're specifically wanting better cargo test reports?

[erratic-pattern]

Made some improvements on my previous version of panic reporting by using actual Diagnostics sources instead of ad-hoc cause chains. So now there's a root Panic report and then a PanicLocation report on top of it if the panic location is available. See code here

Output example with backtrace:

without backtrace:

[zkat]

here's miette's panic hook stuff: https://github.com/zkat/miette/blob/main/src/panic.rs