Re: Re: Challenges around Perl 5.20
On Monday 02 June 2014 17:38:05 Jonas Smedegaard wrote:
> Right, I don't know if cme is limited to atomic statements.
Argh. You're right. cme handles dependencies one-by-one. Even if a dual
statement can be generated, the next cme run will consider them as 2
uncorrelated statements. Then a fix will try to generate a dual statement for
each entry and end up with duplicated dependencies :-(
> Uhm, what are you saying? Do you imply that it means the opposite?
Hum. yes. I don't really like this idea.
> Seems to me that until verified what cme can handle, the question on cme
> support simply remains open. I do believe that above is a valid answer
> to the underlying question on how to possibly express such dependency at
> all.
It's technically correct but looks to me as a trick to work around a dpkg
limitation.
>
> >> - if we would want it instead of an unconditional
> >>
> >> libmodules-build-perl >= x
> >
> > It is much simpler to handle for us and dpkg resolver. Jonas' trick
> > will probably trigger questions from perl packagers (present and
> > futures)
>
> Could you elaborate on which kinds of questions if would trigger to
> express the complex dependency as I propose?
Gregoa did successfully express my concerns.
> Even if not possible for cme to automate what I propose, some (including
> me) might still express it by hand (e.g. to ease backporting), and it
> would be quite relevant to understand what might backfire in that
> approach.
>
> > The only drawback I see: some systems may needlessly install modules
> > when installing backported perl packages.
> >
> > Is that a big deal ?
>
> Becomes bigger if the module needlessly pulled in also need backporting.
> If then the needlessly pulled in and backporting-requiring package has
> sloppy dependencies on e.g. a newer than really needed debhelper or
> Moose, it becomes even bigger...
True. But how many times will that be a problem, say, in the next 3 years ?
> Generally, support at all for backporting is arguably never a big deal.
> Or always. Question is subjective...
Agreed.
All the best
Reply to: