| Index: chrome/chrome_tests.gypi
|
| ===================================================================
|
| --- chrome/chrome_tests.gypi (revision 48246)
|
| +++ chrome/chrome_tests.gypi (working copy)
|
| @@ -2420,7 +2420,23 @@
|
| # src/chrome.
|
| 'message': 'Compiling coverage bundles.',
|
| # MSVS must have an input file and an output file.
|
| - 'inputs': [ '<@(_dependencies)' ],
|
| + #
|
| + # TODO(jrg):
|
| + # Technically I want inputs to be the list of
|
| + # executables created in <@(_dependencies) but use of
|
| + # that variable lists the dep by dep name, not their
|
| + # output executable name.
|
| + # Is there a better way to force this action to run, always?
|
| + #
|
| + # If a test bundle is added to this coverage_build target it
|
| + # necessarily means this file (chrome_tests.gypi) is changed,
|
| + # so the action is run (coverage_bundles.py is generated).
|
| + # Exceptions to that rule are theoretically possible
|
| + # (e.g. re-gyp with a GYP_DEFINES set).
|
| + # Else it's the same list of bundles as last time. They are
|
| + # built (since on the deps list) but the action may not run.
|
| + # For now, things work, but it's less than ideal.
|
| + 'inputs': [ 'chrome_tests.gypi' ],
|
| 'outputs': [ '<(PRODUCT_DIR)/coverage_bundles.py' ],
|
| 'action_name': 'coverage_build',
|
| 'action': [ 'python', '-c',
|
|
|