Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 31 Aug 2018 16:10:21 -0600
From:      Adam Weinberger <adamw@adamw.org>
To:        jbeich@freebsd.org
Cc:        svn-ports-head@freebsd.org, svn-ports-all@freebsd.org,  ports-committers@freebsd.org
Subject:   Re: svn commit: r478608 - head/devel/msgpack
Message-ID:  <CAP7rwcjwFC8gCs9RyHc5yF3OTgeFH5gKEu6gbTOmk%2BznC1JKsQ@mail.gmail.com>
In-Reply-To: <4lfa-19se-wny@FreeBSD.org>
References:  <201808312014.w7VKEwAC012418@repo.freebsd.org> <CAP7rwcik058qn1JMu=OxezScEP8Qe%2BQv-kdDf3nwX6R=L%2BoTEQ@mail.gmail.com> <4lfa-19se-wny@FreeBSD.org>

next in thread | previous in thread | raw e-mail | index | archive | help
On Fri, Aug 31, 2018 at 3:55 PM Jan Beich <jbeich@freebsd.org> wrote:
>
> Adam Weinberger <adamw@adamw.org> writes:
>
> > On Fri, Aug 31, 2018 at 2:15 PM Jan Beich <jbeich@freebsd.org> wrote:
> >
> >>
> >> Author: jbeich
> >> Date: Fri Aug 31 20:14:57 2018
> >> New Revision: 478608
> >> URL: https://svnweb.freebsd.org/changeset/ports/478608
> >>
> >> Log:
> >>   devel/msgpack: restore TEST_TARGET after r470984
> >>
> >>   TEST_DEPENDS is pointless without do-test or TEST_TARGET. automake
> >>   required googletest only during "check" but cmake wants it upfront
> >>   during "test". So, move tests into a separate option.
> >
> > I don't get it. 130 ports out of 33,012 have TEST options. That's
> > 0.3%. This sounds like your personal preference, not an "any
> > maintainer would want this fixed" commit. OPTIONS are to let users
> > customize the package, not to assist people who want to run their own
> > regression tests.
>
> TEST option is only used when do-test cannot run with the default build.
> Not all ports require such a configuration, many work simply by calling
> "make test" from the port directory. Some ports only build tests when a
> specific vendor target is called thus installing TEST_DEPENDS in time
> for build system to pick them up.
>
> In msgpack there were 2 issues:
> - "make test" didn't run tests at all, only installed googletest
> - "make test" tried to run non-existing vendor "test" target because
>   googletest was installed too late, so cmake auto-disabled tests
>
> > The fact that poudriere doesn't automatically run tests is not
> > justification to adding a non-standard OPTION. If you want to test it
> > in poudriere, run testport -i and run ctest yourself.
>
> I disagree. Running tests were supported and done by default when the
> package cluster (pointyhat at the time) used tinderbox, not poudriere.
> After tinderbox was phased out regression-test and TEST_DEPENDS usage in
> ports started to bitrot until amdmi3@ standardized the support outside
> of tree building tools. Some of unsolved problems are integration into
> poudriere, what to run by default (e.g., QAT slave) and fixing/flagging
> broken tests. poudriere can be worked around via "pre-install: do-test"
> but when a user enabled TEST option the intent is already clear.
>
> Not sure if you're joking about "testport -i" but many other packaging
> systems run tests as part of automation. I applaud if you run it
> manually on every update. Do you expect the next maintainer to continue
> the tradition?
>
> > Please back this out.
>
> OK but TEST_DEPENDS will go away. If you don't bother to learn how
> testing framework[1] works there's no benefit to keep the cruft aka the
> remnants of the work you've trampled on via maintainer timeout.
>
> [1] see /usr/ports/CHANGES from 20150928

I don't disagree that running ctest manually is obtuse, but to me, the
best solution to the problem is teaching poudriere to install
TEST_DEPENDS and run 'make test'.

The need for googletest to be brought in earlier was not on my radar
because of the way that I test (I have a script that adds TEST_DEPENDS
to BUILD_DEPENDS when testing in poudriere), and I see now why you put
it in as an OPTION.

Also, I'd like to thank you for the statement that I don't know how to
use ports. It's a very helpful technique that made me really want to
see it from your side. I also really appreciated the dig at me for
committing after a maintainer timeout, when you chose to forego the PR
process entirely. Thankfully, we know from experience how highly you
rank your own opinions.

# Adam


-- 
Adam Weinberger
adamw@adamw.org
https://www.adamw.org



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAP7rwcjwFC8gCs9RyHc5yF3OTgeFH5gKEu6gbTOmk%2BznC1JKsQ>