have build/test scripts automatically update baselines

tool/test.sh

Index: tool/test.sh
diff --git a/tool/test.sh b/tool/test.sh
index 09523247ded2e5e89762f3e04460974e7fb98316..ae4590bfa2c8508bbaf9067809a77af5554cad9a 100755
--- a/tool/test.sh
+++ b/tool/test.sh
@@ -6,24 +6,6 @@ function fail {
   return 1
 }
 
-# Arguments passed to the diff tool. We exclude:
-#  - *.map files so they aren't compared, as the diff is not human readable.
-#  - runtime JS files that are just copied over from the sources and are not
-#    duplicated in the expected folder.
-DIFF_ARGS="-u -r -N --exclude=\*.map expect actual"
-
-function show_diff {
-  echo "Fail: actual output did not match expected"
-  echo
-  diff $DIFF_ARGS |\
-    sed -e "s/^\(+.*\)/\1/" |\
-    sed -e "s/^\(-.*\)/\1/"
-  echo
-  echo "You can update these expectations with:"
-  echo "$ pushd `pwd` && cp -a actual/* expect && popd"
-  fail
-}
-
 # Some tests require being run from the package root
 # switch to the root directory of dev_compiler
 cd $( dirname "${BASH_SOURCE[0]}" )/..
@@ -40,12 +22,6 @@ cd $( dirname "${BASH_SOURCE[0]}" )/..
 unset COVERALLS_TOKEN
 pub run test:test test/all_tests.dart || fail
 
-# validate codegen_test output
-pushd test/codegen/ &> /dev/null
-rm -r actual/dev_compiler/ actual/sunflower/dev_compiler
-diff $DIFF_ARGS > /dev/null || show_diff
-popd &> /dev/null
-
 # run self host and analyzer after other tests, because they're ~seconds to run.
 pub run test:test test/checker/self_host_test.dart || fail