Re: Skip building the test programs in parallel/threadingbuildingblocks

[David Brownlee <abs%absd.org@localhost>]

On Thu, 11 Aug 2022 at 11:11, nia <nia%netbsd.org@localhost> wrote:
>
> On Thu, Aug 11, 2022 at 07:05:04AM +0000, Charlotte Koch wrote:
> > Hi,
> >
> > The CMake in parallel/threadingbuildingblocks says to build the test
> > programs by default. But PKGSRC_RUN_TEST is NOT enabled by default. The end
> > result is: in the default configuration, we're building dozens of
> > executables which aren't even installed.
> >
> > So, I think we can speed up the build significantly in the typical use
> > case by building the test programs iff we want to run the tests in the
> > first place. What do we think of my diff (attached) ?
> >
> > I wonder if there are other, similar opportunities to take advantage of
> > situations like this when PKGSRC_RUN_TEST is disabled. Maybe we could
> > even automate it somewhat.
> >
> > Charlotte
>
> Hiya charlotte,
>
> I agree that PKGSRC_RUN_TEST is likely to be an extremely uncommon
> configuration! Lots of stuff breaks with it.
>
> However, for me at least, as a package maintainer, the most common
> way I run tests is 'make test' after updating a package or making
> some changes, to check there are no regressions. This works even
> if PKGSRC_RUN_TEST is no.
>
> So I'm not sure making test builds depend on PKGSRC_RUN_TEST is a
> good idea. We could have a discussion about making it depend
> on PKG_DEVELOPER instead, which I think would be more reasonable.
> But there's also people who think PKG_DEVELOPER=yes should be
> the default - bulk builds usually run with it to catch problems,
> for example.

Maybe there is scope for moving some of the features currently under
PKG_DEVELOPER into the default behaviour, and adding a
PKGSRC_FAST_AND_LOOSE option (alternate names proposals may be
considered :-p) which disables them. Either way, would agree that
building tests fits nicely under PKG_DEVELOPER.

David