std without backtrace?

[aep]

backtrace is fairly heavyweight and sometimes straight up doesnt work (on mips for example).

Unfortunately the std feature depends on backtrace. Is that just something that needs fixing or is it intentional?

[yoshuawuyts]

I hope I'm getting this right, but I believe backtraces can be toggled using the RUST_BACKTRACE env var which means they should have no performance impact unless enabled.

[aep]

doesnt compile on android either. it refferences a GNU only symbol __errno_location

so both mips and android are broken

[yoshuawuyts]

If it's GNU-only, wouldn't that mean that Windows would have similar hurdles? To my knowledge failure also works on Windows.

Perhaps there's a non-GNU solution available there that might serve as a basis for fixing MIPS/Android?

[aep]

probably not. this is in unix.o (ironic, since thats GNU not unix)

anyway, this is all i need to do to disable backtrace. it compiles fine, apparantly its already behind a feature switch.

is there any way i can do this without forking?

diff --git a/Cargo.toml b/Cargo.toml
index d8c505a..7be6605 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -21,7 +21,7 @@ version = "0.3.3"
 members = [".", "failure_derive"]
 
 [features]
 default = ["std", "derive"]
 #small-error = ["std"]
-std = ["backtrace"]
+std = []
 derive = ["failure_derive"]

[aep]

allright. i have forked failure with the one line fix, since patch doesnt seem like a good permanent solution

https://crates.io/crates/failure_without_backtrace

[vorner]

I could use such thing too.

Maybe remove the std→backtrace dependency, but put backtrace between defaults? Could this be done?