svn commit: r392209 - in head/devel: . p5-Minilla

[Adam Weinberger]

> On 16 Jul, 2015, at 8:30, Alexey Dokuchaev <danfe at FreeBSD.org> wrote:
> 
> On Thu, Jul 16, 2015 at 05:20:44PM +0200, Baptiste Daroussin wrote:
>> On Thu, Jul 16, 2015 at 03:17:30PM +0000, Alexey Dokuchaev wrote:
>>> Fair enough; [...] lets make it easier for new folks to get a hold of
>>> things, no problem.
>>> 
>>> But forbidding a 100% perfectly valid, supported, and documented syntax
>>> used in good will and intention by an experienced developer is IMO wrong.
>> 
>> Given the number of times I had fixed side effect of this I bet I'm would
>> like to know how many people really knows how make(1) works :) in particular
>> in that area.
> 
> When sat@ was actively doing his peer reviews, he would've always point it
> out to folks that misused lazy vs. immediate expansion.  Too bad he's no
> longer active; I think that rather than trying to avoid sharp corners of our
> make(1), or any tool FWIW, we should be doing continuous education work akin
> to what he did...
> 
>> With the new way of validating dependencies (where the pattern is now also
>> checked after the dependency is installed, people would now quickly noticed
>> but before this recent change it was most of the time silent. poudriere was
>> the only tool able to notice that).
> 
> This actually brings us to another rule which makes more sense to become a
> "hard" one: DO NOT commit ports work without poudriere/tinderbox test run.

I’d just like to mention PR 191273, which suggesting adding a new depends macro
that registers a BUILD_ and RUN_DEPENDS.

	BUILDANDRUN_DEPENDS=	foo:...
(Or something. I’m terrible at naming things, but the concept/patch is there.)

By way of examples, a huge percentage of Perl modules rely on :=, and many of
them perform impressive manoeuvres to assign
	RUN_DEPENDS:=	${BUILD_DEPENDS}
	BUILD_DEPENDS+=	…
or my favorite
	BUILD_DEPENDS=	foo:...
	RUN_DEPENDS:=	${BUILD_DEPENDS:Nfoo*}

The fact that not even the PHB was clear on whether := was correct suggests that
there is a problem that can use a solution.

# Adam


—
Adam Weinberger
adamw at adamw.org
http://www.adamw.org