Remove -bench from ./ci/test.sh

Testing should not run benchmarks by default, it takes too long.

ci/test.sh

@@ -10,18 +10,16 @@ argv=(
   "--junitfile=ci/out/websocket/testReport.xml"
   "--format=short-verbose"
   --
-  -race
   "-vet=off"
-  "-bench=."
 )
-# Interactive usage probably does not want to enable benchmarks, race detection
-# turn off vet or use gotestsum by default.
+# Interactive usage does not want to turn off vet or use gotestsum by default.
 if [[ $# -gt 0 ]]; then
   argv=(go test "$@")
 fi
 
-# We always want coverage.
+# We always want coverage and race detection.
 argv+=(
+  -race
   "-coverprofile=ci/out/coverage.prof"
   "-coverpkg=./..."
 )

docs/CONTRIBUTING.md

@@ -43,12 +43,10 @@ For coverage details locally, please see `ci/out/coverage.html` after running `c
 
 See [ci/image/Dockerfile](ci/image/Dockerfile) for the installation of the CI dependencies on Ubuntu.
 
-You can also run tests normally with `go test`.
-`ci/test.sh` just passes a default set of flags to `go test` to collect coverage,
-enable the race detector, run benchmarks and also prettifies the output.
+You can also run tests normally with `go test`. `ci/test.sh` just passes a default set of flags to
+`go test` to collect coverage, enable the race detector and also prettifies the output.
 
-If you pass flags to `ci/test.sh`, it will pass those flags directly to `go test` but will also
-collect coverage for you. This is nice for when you don't want to wait for benchmarks
-or the race detector but want to have coverage.
+You can pass flags to `ci/test.sh` if you want to run a specific test or otherwise
+control the behaviour of `go test`.
 
 Coverage percentage from codecov and the CI scripts will be different because they are calculated differently.