274be6de55
This is a re-factor of element_dependencies to achieve two things -- centralising override policy and storing path names. Firstly we want to make the override policy for elements completely explicit. Currently, elements that wish to copy parts of other elements walk ELEMENTS_PATH themselves and look for elements in IMAGE_ELEMENT. How they handle duplicate elements can differ, leading to inconsistent behaviour. We introduce logic in element-info to find elements in each of the directories in ELEMENT_PATHS in *reverse* order -- that is to say, earlier entries in the paths will overwrite later ones. For example ELEMENT_PATHS=foo:bar:baz will mean that "foo/element" will override "baz/element", since "foo" is first. This should be sane to anyone familiar with $PATH. Documentation is clarified around this point and a test-case is added. The second thing is that we want to keep the complete path of the elements we have chosen. We want the aforementioned elements that walk the element list to use these canonical paths to pickup files; this way they don't need to make local decisions about element overrides, but can simply iterate a list and copy/merge files if they exist. A follow-on change (I7092e1845942f249175933d67ab121188f3511fd) will expose this data in a separate variable that can be parsed by elements (a further follow-on I0a64b45e9f2cfa28e84b2859d76b065a6c4590f0 modifies the elements to use this information). Thus this does not change the status-quo -- elements that are walking ELEMENTS_PATH themselves and can/will continue doing that. Change-Id: I2a29861c67de2d25c595cb35d850e92807d26ac6 |
||
---|---|---|
.. | ||
__init__.py | ||
base.py | ||
test_elementdeps.py | ||
test_elements.py | ||
test_loggingconfig.py | ||
test_no_dup_filenames.py |