Introduce a proper error handling framework

[rust.git] / ci.sh

diff --git a/ci.sh b/ci.sh
index b914477d44a81ca5817d9818d54b83f807d7868f..e7c6ed833c60c4b60bed064f5128570375a5f70b 100755 (executable)
--- a/ci.sh
+++ b/ci.sh
@@ -1,15 +1,16 @@
 #!/bin/bash
 set -euo pipefail
+set -x
 
 # Determine configuration
 export RUSTFLAGS="-D warnings"
 export CARGO_INCREMENTAL=0
-export CARGO_EXTRA_FLAGS="--all-features"
+export CARGO_EXTRA_FLAGS="--all-features" # in particular, expensive-debug-assertions
 
 # Prepare
 echo "Build and install miri"
-./miri build --all-targets --locked
-./miri install # implicitly locked
+CARGO_EXTRA_FLAGS="" ./miri install # implicitly locked -- and the *installed* Miri does *not* get the expensive-debug-assertions feature
+./miri build --all-targets --locked # the build that all the `./miri test` below will use
 echo
 
 # Test
@@ -40,6 +41,13 @@ function run_tests {
   # any interactive questions.
   ${PYTHON} test-cargo-miri/run-test.py
   echo
+
+  # Ensure that our benchmarks all work, on the host at least.
+  if [ -z "${MIRI_TEST_TARGET+exists}" ]; then
+    for BENCH in $(ls "bench-cargo-miri"); do
+      cargo miri run --manifest-path bench-cargo-miri/$BENCH/Cargo.toml
+    done
+  fi
 }
 
 function run_tests_minimal {