have build/test scripts automatically update baselines

tool/test.sh

 #!/bin/bash
 set -e # bail on error
 
 function fail {
   echo -e "Some tests failed"
   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]}" )/..
 
 # Check minimum SDK version
 ./tool/sdk_version_check.dart 1.9.0-dev.4.0 || fail
 
 # Make sure we don't run tests in code coverage mode.
 # this will cause us to generate files that are not part of the baseline
 # TODO(jmesserly): we should move diff into Dart code, so we don't need to
 # worry about this. Also if we're in code coverage mode, we should avoid running
 # all_tests twice. Finally self_host_test is not currently being tracked by
 # code coverage.
 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
 
 {
   fc=`find test -name "*.dart" |\
       xargs grep "/\*\S* should be \S*\*/" | wc -l`
   echo "There are" $fc "tests marked as known failures."
 }
 
 echo -e "All tests pass"