Description
Stack traces would be much easier to digest if the lines from stdlib were removed before being printed to the user. For example, take this output from calling option.unwrap on a None value in a test from a project I'm working on:
---- compare_and_delete stdout ----
thread 'compare_and_delete' panicked at 'called `Option::unwrap()` on a `None` value', /home/rustbuild/src/rust-buildbot/slave/stable-dist-rustc-linux/build/src/libcore/option.rs:362
stack backtrace:
1: 0x7f12357e3689 - sys::backtrace::write::hbc46dc0cfb3b9537d4r
2: 0x7f12357e6c46 - panicking::on_panic::h74d3c14d86c58ac8jrw
3: 0x7f12357d8a92 - rt::unwind::begin_unwind_inner::h382cea404b11eb00t6v
4: 0x7f12357d94ec - rt::unwind::begin_unwind_fmt::h5c14cfc30901d9d274v
5: 0x7f12357e68d6 - rust_begin_unwind
6: 0x7f1235814184 - panicking::panic_fmt::h9ea3571c30b632b3wwy
7: 0x7f1235814104 - panicking::panic::h0265a6efa68361e53uy
8: 0x7f1235660991 - option::Option<T>::unwrap::h8561976572633329008
at /home/rustbuild/src/rust-buildbot/slave/stable-dist-rustc-linux/build/src/libcore/macros.rs:19
9: 0x7f12356612c4 - compare_and_delete::h6f5ab680c2a34c3feia
at tests/client_test.rs:62
10: 0x7f123575e39b - boxed::F.FnBox<A>::call_box::h15853468894108720310
11: 0x7f123576190b - boxed::F.FnBox<A>::call_box::h15343356339842678509
12: 0x7f123575eb59 - rt::unwind::try::try_fn::h11678192874774444076
13: 0x7f12357eb6a8 - rust_try_inner
14: 0x7f12357eb695 - rust_try
15: 0x7f123575ee08 - boxed::F.FnBox<A>::call_box::h13152174250445044076
16: 0x7f12357e5c21 - sys::thread::create::thread_start::h490278b5c3c0b49faqv
17: 0x7f12340ed0a3 - start_thread
18: 0x7f12348ed04c - clone
19: 0x0 - <unknown>
While it may be accurate, what I want as a programmer is to know what part of my code caused the exception. The error message only tells me about the implementation of the Option type, and I have to dig through a stack trace of stuff from the stdlib to find the line that tells me it came from tests/client_test.rs:62 in my code.
I talked with @wycats about this briefly at the Rust 1.0 launch party and he agreed that we probably need something akin to ActiveSupport::BacktraceCleaner from Rails.
This might actually be two issues – one is that the error message should show the line in my code where the unwrap happened, not the line in option.rs where the panic was actually triggered, and the other is the filtering of lines in the stack trace. Let me know if I should split part of this off into a separate issue.