feat: enable backtraces by default

This is a great help in debugging, and I don't think there's much
downside to doing that? Capturing a backtrace is somewhat costly,
especially on windows, but given that we run with panic=abort, that's
bound to be a pretty cold path anyways!

To disable backtraces, set `RUST_BACKTRACE=0` env variable.

CHANGELOG.md

@@ -23,6 +23,9 @@
   Instead of it aggregate `near_peer_message_received_by_type_total` metric
   instead.  For example, to get total rate of received messages use
   `sum(rate(near_peer_message_received_by_type_total{...}[5m]))`.
+* Backtraces on panics are enabled by default, so you no longer need to set
+  `RUST_BACKTRACE=1` environmental variable. To disable backtraces, set
+  `RUST_BACKTRACE=0`.
 
 ## 1.28.0 [2022-07-27]
 

neard/src/main.rs

@@ -1,13 +1,13 @@
 mod cli;
 mod log_config_watcher;
 
-use near_primitives::version::{Version, PROTOCOL_VERSION};
-use near_store::version::DB_VERSION;
-
 use self::cli::NeardCmd;
 use crate::cli::RunError;
+use near_primitives::version::{Version, PROTOCOL_VERSION};
+use near_store::version::DB_VERSION;
 use nearcore::get_default_home;
 use once_cell::sync::Lazy;
+use std::env;
 use std::path::PathBuf;
 use std::time::Duration;
 
@@ -42,6 +42,11 @@ static ALLOC: near_rust_allocator_proxy::ProxyAllocator<tikv_jemallocator::Jemal
 static ALLOC: tikv_jemallocator::Jemalloc = tikv_jemallocator::Jemalloc;
 
 fn main() -> Result<(), RunError> {
+    if env::var("RUST_BACKTRACE").is_err() {
+        // Enable backtraces on panics by default.
+        env::set_var("RUST_BACKTRACE", "1");
+    }
+
     rayon::ThreadPoolBuilder::new()
         .stack_size(8 * 1024 * 1024)
         .build_global()