Re: build with code coverage

On Tue, Jul 9, 2013 at 11:06 PM, Michael Stahl <mstahl@redhat.com> wrote:
> On 09/07/13 22:04, John Smith wrote:
> > On Tue, Jul 9, 2013 at 9:49 PM, bfo <bfo.bugmail@spamgourmet.com> wrote:
> > > John Smith wrote
> > > > I also ran an lcov report on current master, and uploaded the results
> > > > here:
> > > > http://dev-builds.libreoffice.org/lcov_reports/master~2013-07-08_14.53.51/
> > >
> > > H!
> > > Thanks a lot, hope you can run it and upload at least once after branching.
> > > Anyone can comment the results vs master~2012-12-23_14.37.59? Are those
> > > comparable?
> >
> > Remind me later on this list after your branch, and ill see what i can do.
> >
> > Comparing earlier results, the amount of code lines 'hit' by the
> > tests/checks seems to have increased a lot, so I guess that is a good
> > thing.
> >
> > But if you really want comparisons to have any siginificant meaning, I
> > guess some serious time needs to be invested in :
> >
> > 1.) improving and expanding the tests.
> > 2.) running the lcov/gcov code coverage at regular intervals.
> >
> > When posting on this list, I just get the impresssion that there just
> > isnt enough interest at this point to realize that right now. So I
> > guess that right now generating the reports is primailry for my
> > personal educational value, which is just fine by me; im learning as I
> > go.
>
> i'm afraid that right now we have so many obvious large gaps in test
> coverage that using test coverage tools is a bit premature.  perhaps one
> run per release cycle would be informative for general trends but likely
> few developers have the time to look at the result in detail anyway (i
> don't).
Yes, of course this is what i meant to say.
;)



Regards,

John Smith.