On Sat, Apr 14, 2007 at 12:51:36PM +0200, "Andreas J. Koenig" <[EMAIL
PROTECTED]> wrote:
> > The bug seems obviously to be a missing dependency in Continuity: If it
> > requires Event, it needs to specify that in its dependencies explicitly,
> > as Coro (quite correctly) does not depend on Event at all.
>
> You are shooting too quickly. Continuity *does* specify a dependency
> on Coro::Event.
Great. And Coro::Event depends on Event. That looks just fine to me.
Besides, you cannot use Coro::Event without Event. Any program that
uses Coro::Event does depend on Event. Even if a module does not itself
require Event, the main program does, so the dependency is always there
_directly_.
> Coro::Event says the following during Makefile.PL execution (I did
> cite this sentence in the previous posting):
>
> Warning: prerequisite AnyEvent 2.51 not found.
> Warning: prerequisite Event 1.06 not found.
> Warning: prerequisite IO::AIO 2.3 not found.
>
> *** Event not found, not build Event support.
>
> This is a bug in my eyes.
Not in mine. Coro::Event is an optional part of Coro. Not building it when
the user fails to install it is obviously a fetaure, not a bug.
> It does NOT let the user specify that he really wants
> Coro::Event. Except by *installing* something else beforehand. You are
> suggesting that this is the way things *should* be done?
Given that the rest of the world does it that way, it at least is the
expected way. If build other software that has optional parts requiring
prerequisites and they are not installed, they do not build it. Thats why it
is optional.
At least on unix, that is the prevalent philosophy...
Coro (or actually ExtUtils::MakeMaker) loudly complains about missing
prerequisutes. Besides that Coro has a hard dependency on Event currently,
which is a bug, having optional dependencies is a feature.
> > Forcing Event for no reason seems like a very bad idea - if a package
> > relies on Event it should have that dependency, adding it to Coro to help
> > those broken packages makes little sense, IMnsHO.
>
> You are so right, but I did not ask you to "Forcing Event for no
> reason" but I said:
>
> Ideally, Marc would ask the user if he wants to build Event
>
> I correct myself now, I wanted to say
>
> Ideally, Marc would ask the user if he wants to build Coro::Event
> ^^^^^^
That "ideally" is a large burden on users who can read, or use CPAN. I
do not see why I should make configurability for users even more of a
nuisance than it already is just because some other modules (Continuity)
have dependency bugs.
> If my sloppiness has lead you to a wrong impression, I'm sorry.
Not really, I just do not see the point.
> > The result would only be that Coro would fail to work if Event fails to
> > work, even though Coro itself doesn't require that (and a large class of
> > Coro users do not require Event either). This seems like a bad tradeoff to
> > what seems essentially a workaround for some other module's bug.
>
> If the user has expressed the wish, so be it.
Exactly, this is the current situation. Just the mechanism is less of a
nuisance.
> > As such, I would be reluctant to hack around in Coro to add a fake
> > dependency that Coro does not really require.
>
> I presume the above correction of my wish sets straight what I'm
> really suggesting?
I never assumed otherwise. After all, you cannot command me, and I found
that my expressed wishes are not always granted, either, so it would only be
logical to assume it vice versa.
> I specifically do not ask you to add a fake dependency that the user
> does not require. I ask you to do the opposite: let the user who
> requires it, build Coro::Event.
That is the current situation which goes back till the very dark ages of
Coro. So obviously, you are asking for something different.
--
The choice of a
-----==- _GNU_
----==-- _ generation Marc Lehmann
---==---(_)__ __ ____ __ [EMAIL PROTECTED]
--==---/ / _ \/ // /\ \/ / http://schmorp.de/
-=====/_/_//_/\_,_/ /_/\_\ XX11-RIPE
