[macports-mgr] Dependencies and variants

[C. Florian Ebeling]

On Thu, Nov 6, 2008 at 11:27 AM, Paul Guyot <pguyot at kallisys.net> wrote:
>
> Le 6 nov. 08 à 10:44, Ryan Schmidt a écrit :
>
>> I would still like this discussion to take place on macports-dev, not
>> macports-mgr.
>
> True.

Why are such things discussed on a closed list?! I was already
wondering why there is no discussion going on. Isn't this projects
also built on the notion of openness etc? Also I didn't hear
anything about the R-word since the the management is
in office.

I think I'm a patient person but mp starts to bore me abit tbh.


>
> A quick summary for macports-dev: the discussion is about r41526 and r41527,
> i.e. allowing a new syntax for dependencies in the next release without
> actually implementing them. The idea is to allow work and testing for bug
> #126 (dependencies should take variants into consideration) on trunk, with
> ports in trunk using the new syntax but remaining compatible with released
> base/. Experience with livecheck showed that if we wait for the
> implementation to support the syntax, we will need several releases to get
> it right and get the new syntax adopted by maintainers. I asked portmgr to
> consider it for the next release, and portmgr asked what I had in mind for
> the syntax, since there is just no implementation or guideline in r41526 and
> r41527. To what I replied that I prefer not to force any syntax for many
> syntaxes came up in the past for #126, and my belief is that it's eventually
> up to whoever implements that. Therefore, the changes allow for :
>
> depends_*       port:[whatever:]port_name
>
> where we currently have
>
> depends_*       port:port_name
>
> as this is coherent with the current depends_* lines.
>
> It also allow for a new syntax
>
> depends         whatever
>
> if whoever works on #126 wants to get rid of the old depends_* syntax for
> more freedom. While I would just be happy with whatever comes up that fixes
> #126, I argued that depends_* were somewhat ugly and confusing.
>
>> I don't see why depends_* is ugly or should be abandoned.
>
> :)
>
>> depends_build is for things which are needed to build the software. Note:
>> this is not only the build phase, but (in MacPorts trunk) any phase involved
>> in building the software, including fetch, extract, and configure. (I think
>> I'm remembering that right.)
>
> AFAIK, no dependency is processed for extract and fetch. In other words, all
> ports that use fetching using git on 10.5 and fetching using svn and git on
> 10.4 just don't fetch with trace mode enabled.
>
>> Things declared as depends_build can be uninstalled after the port is
>> installed with no ill effects. The others cannot. Putting all dependencies
>> into a single "depends" negates this possibility.
>
> Depends how the depends syntax is designed. It could be:
>
> depends {
>        build { awk }
>        lib { glib2 }
>        run { evince }
> }
>
> Paul
>
> _______________________________________________
> macports-dev mailing list
> macports-dev at lists.macosforge.org
> http://lists.macosforge.org/mailman/listinfo.cgi/macports-dev
>



-- 
Florian Ebeling
Twitter: febeling
florian.ebeling at gmail.com