Bug 38840 - Adding coverage analysis to unit tests
mstahl at redhat.com
Thu Aug 23 11:41:43 PDT 2012
On 23/08/12 18:09, John Smith wrote:
> I just finished a first full run of lcov. There was one 'make check'
> failure though, and there were a lot of 'warnings' running lcov that
> may need some further investigation. Also, there is some stuff
> included ('/usr/include/boost', for example) that might not be desired
> in the report ?
yes, everything under /usr/include is just noise, would be great to
filter that out.
hmm... it's a bit ugly that we have so many directories in solver that
headers are copied to, but i guess we have to live with that for now :)
> I guess the main thing to do first now is to see if this report
> actually makes any sense. Essentially, all code that gets executed by
> 'make check' on toplevel (which does dev-install and subsequentcheck,
> and dev-install includes both unitcheck and slowcheck) should show up
> as covered in the report. Maybe people that are familiar with the
> contents of the checks/tests and what code/functionality they cover
> can take a look at that ?
so looking at a few bits where i'm familiar with the tests tests like:
it seems quite plausible.
but looking at
some functions that are definitely tested like
librdf_Repository::querySelect or CNode::insertBeforeshow up entirely
are you running the JUnit based tests as well? i.e. if you use
--without-java or --without-junit, that would negatively affect the test
these are run during "make subsequentcheck", you should have files like
> Anyway, the generated html report as it currently is can be found here :
very good work!
More information about the LibreOffice