telemetry mode hangs if there are too many errors

[arielb1]

Meta

$ rustup -V
rustup 0.5.0 (4be1012 2016-07-30)
$ rustc -V
rustc 1.12.0-nightly (0a3180baa 2016-08-03)
$ cat ~/.multirust/settings.toml 
default_host_triple = "x86_64-unknown-linux-gnu"
default_toolchain = "nightly"
telemetry = true 
version = "12"

[overrides]

STR

$ for i in $(seq 1 10000); do echo "fn fail() {}"; done >> so-much-fail.rs
$ rustc so-much-fail.rs 

Expected Result

10000 errors are printed to standard error

Actual Result

hang, since telemetry_rustc (https://github.com/rust-lang-nursery/rustup.rs/blob/4be1012/src/rustup/command.rs#L51) pipes stderr and waits, without reading from it.

moved from rust-lang/rust#34535

[brson]

Thanks for filing this. Telemetry is very underbaked right now and I don't recommend anybody use it.

[aidanhs]

Just because it could break things or because it might accidentally be collecting all of my hard drive? If the former that's fine, in general I feel capable of debugging issues if I hit them (though @arielb1 saved me for this one).

[TimNN]

Just ran into this, very annoying to debug, although I did learn a bit more about lldb.