Docs talk about limitations of backtrace in the whatever type based on backtrace not being stable yet, but backtrace has now been stabilized

[Frederik-Baetens]

Docs state:

When wrapping errors, only the backtrace from the shallowest function is guaranteed to be available. If you need the deepest possible trace, consider creating a custom error type and using #[snafu(backtrace)] on the source field. If a best-effort attempt is sufficient, see the backtrace method.

When the standard library stabilizes backtrace support, this behavior may change.

But backtrace has since been stabilized: rust-lang/rust#99573

https://docs.rs/snafu/latest/snafu/struct.Whatever.html#limitations

I'm raising this issue to inquire about what this could mean for snafu. Does this enable improvements to snafu's backtrace support? Or should things stay as they are with only a minor tweak to the docs language?

If you could provide some guidance on how the stabilization of backtrace could improve snafu, I'd love to take a crack at implementing that improvement.

[shepmaster]

When I wrote those docs, I had assumed that the Backtrace type would stabilize at the same time as the Error::backtrace method as they were the same feature gate.

Since then, the two were detangled from each other and then the backtrace method fully removed in favor of the provide API, which is still unstable.

There's still no stable, standard-library general way of querying if an arbitrary error contains a backtrace or not, which is really what SNAFU needs to have better user experience here. Once there is, we can have something that walks the source chain and returns the deepest Backtrace, which will give the most information. We can also check to see if a nested error already has a backtrace and if not, capture one of our own, avoiding duplicate work in some cases.

For now, there's not much to do, but perhaps the docs could be tweaked to distill some of this information.

[Enet4]

Related issues:

• Embedding a whatever broken on nightly with unstable-provider-api #373: "Only reason I needed (to use nightly) it was that it wouldn't print backtrace otherwise. I don't want to use unstable features."
• The path to 1.0 #132. I updated it to reflect that Backtrace is now available in std but we still need the provider API to be stabilized before we can have SNAFU 1.0.