[cdt-dev] Aboud content assist failures and committing failing Junit tes

["Ploett, Norbert" <norbert.ploett@xxxxxxxxxxx>]

Hi folks,

so I ported a number of Junits for the content assist processor over to
the new CCompletionProcessor2 and they generally fall into 2 categories:

1) After the port the tests can be made to work reasonably, there are
minor deviations from the expected results.

2) The completion processor cannot create a completion node at all.

For category 1) I commited these tests to the repository, they can be
found in package org.eclipse.cdt.ui.tests.text.contentassist2. I made
verbose comments, describing the deviations, but twisted the code so
that the tests shall be green when they are run. The clean way would be
to let the tests be red and create bugzilla entries describing the
problems. But I was shy to commit test code that would fail.

For category 2) I did not yet commit these because they fail so badly,
and, as above, I did not want to commit failing tests. I do not think
that I will be able to repair the deficiencies on my own, for the moment
I can just be the supplier of well-documented bug reports.


Here are some alternatives of what I could do next:

Category 1 failures:
a) Create bugzillas for the category 1 failures, but leave the code so
that the tests remain green.
b) Create bugzillas for the category 1 failures, and make the tests
actually fail until the completion processor and it's subsidaries are
repaired.

Category 2 failures:
a) Create bugzillas for the category 2 failures, with the test case code
attached, but not commit the code to the repository.
b) Create bugzillas for the category 2 failures, and commit them to the
repository.

... and the final big question: Should the tests in
org.eclipse.cdt.ui.tests, including my content assist stuff, be added to
the nightly build tests? (See also my earlier email of today for this
subject)

Thanks for opinions


Norbert Ploett