[cmake-developers] Support for coverage.py coverage on the topic stage

Bill Hoffman bill.hoffman at kitware.com
Mon Sep 30 14:44:52 EDT 2013


On 9/30/2013 11:24 AM, Rolf Eike Beer wrote:
>
> For all other handlers it may happen that more than one coverage db is
> found in the build tree and they would both be used. This is really
> useful, e.g. if you want to have coverage for both C and Python code.
> This will not work with the BullsEye handler involved. I think it would
> be best if that would be refactored to have the same return values as
> the other handlers. The file_count isn't really needed as file_count in
> cmCTestCoverageHandler::ProcessHandler(), at least not in the first
> half. It's just a marker "has anything been processed at all" there, so
> it doesn't matter if the bullseye method returns 1 or the correct file
> count there.

OK, I took a look at the code and the problem is in how ProcessHandler 
works.  For the other handlers that keep track of lines covered and not 
branches, it does some processing at the end.  That processing makes no 
sense for the branch based coverage of Bullseye.   So, when Bullseye 
works it does not do any of the other summary stuff at the bottom nor do 
we want it too.   It returns in that case.   I had not considered the 
possibility of a project with more than one coverage thing going on at 
the same time.  Certainly it would not make sense to do two different 
c++ coverages at the same time.  However, you might want python or some 
other language.

-Bill





More information about the cmake-developers mailing list