On Wed, Feb 28, 2018 at 11:56:07AM +0100, Lukáš Hrázký wrote:
> On Tue, 2018-02-27 at 18:02 +0100, Christophe Fergeau wrote:
> > On Tue, Feb 27, 2018 at 01:04:31PM +0100, Lukáš Hrázký wrote:
> > > On Tue, 2018-02-27 at 06:33 -0500, Frediano Ziglio wrote:
> > > > >
> > > > > Removing noinst_PROGRAMS from src/unittests/Makefile.am will cause the
> > > > > unit tests not be built on a regular build (`make`), they will only be
> > > > > built when running the tests (`make check`).
> > > > >
> > > > > C++ unit test frameworks are notorious for long build times too.
> > > > >
> > > > > Signed-off-by: Lukáš Hrázký <lhrazky@xxxxxxxxxx>
> > > > > ---
> > > > > I am not sure if there was another reason for noinst_PROGRAMS, but I
> > > > > don't think so.
> > > > >
> > > >
> > > > Yes, this is mainly what it does. In spice-server we have same pattern,
> > > > the reason to have it is the opposite of this patch, we want to make
> > > > sure tests compile even if they are not executed.
> > > >
> > > > If the final aim of this patch is to avoid the requirement of catch
> > > > library this patch is just partial.
> > > >
> > > > If the aim is to just save some time avoiding to compile tests
> > > > the patch is fine but looks like presenting different behaviour
> > > > comparing to other SPICE projects.
> > >
> > > Not to avoid the Catch dependency, to save time and just not do what
> > > the user doesn't want to do. If he wants to build the project, he
> > > doesn't need the tests. You only really need to build the tests if you
> > > want to run them.
> > >
> > > For SPICE projects in general, I think the logic is backwards. We do
> > > not want to build the tests on a regular build to make sure they do
> > > build. We want to (build and) run the unit tests all the time to make
> > > sure not only they build, but also that they pass :)
> >
> > This logic is following what the autotools give us, a separate make
> > check target.
>
> I am not following :) I think you could make the default (build) target
> depend on the check target (practically making them one target, as
> opposed to a separate check target, which you mentioned), if that is
> what you mean. But that makes little sense, as often you want to only
> build and not run tests.
All I meant is that when you use the autotools, what you get by default
is a make (all) target, and a make check target, which are separate. So
I was saying that your suggestion (always running make check as part of
make (all)) is not what the autotools offer by default, which is why
it's not done this way at the moment. I was not saying it's not possible
to do it ;)
>
> > With this patch, the situation becomes odd I think,
> > configure.ac requires an unusual (imo) dependency ('catch') which is
> > then not going to be used in our default build (a straight 'make').
>
> How is it more odd than having an unusual dependency which is only used
> to build the tests which are then not run? :) I find it more odd to
> build the tests for no reason (other than to make sure they build, but
> the 'make sure' part is only our responsibility and makes no sense for
> users that just want to use it).
Hmm, the last part of your sentence could apply to unit tests too ;)
>
> By the way, I only sent this change, because it makes sense to me in
> itself, but I am somewhat expecting someone (who has more configure.ac-
> fu than me) to post the part that makes Catch an optional dependency...
Ah, I did not realize you needed a hand with that, which is why I did
not propose any patch.
Christophe
>
> Cheers,
> Lukas
>
> > Christophe
Attachment:
signature.asc
Description: PGP signature
_______________________________________________ Spice-devel mailing list Spice-devel@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/spice-devel