Hi all,
(responses embedded below)
> On 15 Nov 2010, at 13:06, Charles Lepple wrote:
>
>> On Nov 14, 2010, at 11:57 PM, David Fang wrote:
>>
>>> There isn't anything we can do about Apple's decision, unfortunately.
>>> Unless you actually intend to use the java interface for ppl, it
>>> shouldn't
>>> be an issue if this package drops the Java module in the future.
>>
>> I haven't checked to see if any Fink packages use the Java module, but
>> Fink's dependency engine seems to be better about handling cases where
>> a single .info file with splitoffs is converted into two or more .info
>> files (carve-offs?), each with a reduced set of BuildDepends.
>>
>> This way, the base ppl package wouldn't need to depend on Java, but
>> there would still be a Java option if someone has installed all the
>> external dependencies. The nice thing for end users is that if it's
>> the same source file, they most likely already have it in /sw/src (and
>> they won't need to re-download it for the Java carve-off).
>>
>> Barring any general objections from the list, let me know if you want
>> help testing something like that. I could probably remove the Java
>> headers from one of my machines here and not run into any other
>> problems outside Fink.
I admit there is much that can be improved. This latest iteration was
intended to address the system-java-dev build issue, fill in a few missing
dependencies, and make it work on 10.4. (cloog should be enabled on 10.4
now too.) I just wanted to make it less broken first. (I did not create
the initial packaging for this, so I wanted to minimize perturbation.)
> There are a bunch of problems of that sort with ppl, and it is probably
> best to address all of them at once.
>
> 1) The package builds differently according to the presence or not of
> other fink pkgs. From old notes, and from memory, this includes glpk and
> most prolog pkgs (gprolog, swi-prolog, yap), ocaml and possibly some
> other ocaml- related pkgs (ocaml-lib and/or ocaml-findlib).. Depending
> to the specific versions of ppl itself and/or the above fink pkgs and/or
> the fink installation there may be one or other of the above that fails,
> but the above is the typical picture of what the pkg tries to do.. And
> there may be 1 or 2 of the above that are only triggered by additional
> configure params, but for most this is not the case.
I've addressed glpk in my latest packaging; it is only used in the
testsuite and never linked in the final libs, nor referenced by the
installed includes. Thus, I've listed it in TestDepends. I've noted that
ocaml is checked for in configure and at least used in testing, but
haven't determined whether or not it affects the package: it's on my TODO
list. My 3+ day-long testing of the ppl test suite happened to find ocaml
during configure, so that enabled and covered any ocaml tests. When I get
around to it, I can see what it does with prolog. Would it be safe to
just build-conflict against these as long as they don't change the package
contents?
> 2) As the above shows, this is a pkg intended for a "prolog-type
> audience", (part of the) OR theory community, and the like; and belongs
> to the section "sci". It is most basically and originally an exact
> implementation of Fourier- Motzkin as I remember, plus more generally of
> mixed linear programming, and a bit more. All such pkgs are in the
> section "sci", and (a full version of) ppl belongs there too.
I agree 'sci' is more appropriate, but I simply wanted to fix it in-place
for now. Is there any policy against moving this to a different section?
(By similar argument, gmp, mpfr would belong to 'sci' too; it feels like
the right home for purely mathematical and algorithmic packages.)
> 3) Such a full version should probably be essentially fully enabled, if
> only to respect the upstream intent ("upstream knows better"). On the
> other hand, it is not reasonalbe to let gccXY have such a list of
> dependencies. This would call for a varianted pkg, with a "mini" variant
> providing just what the gccXY pkgs actually need, and taking great care
> to build identically even if any of the above fink pkgs (or other
> conceivably relevant local pkgs) are installed.
I understand the rationale for doing this. I don't know what the minimal
set of interfaces/features is required for gcc w/ cloog (I doubt the Java
interface is required), perhaps Jack Howarth can comment? Verifying these
variants would take a little more time than what I have now.
> 4) In other not to disrupt existing pkgs, it might be more convenient to
> call the "mini" variant "ppl", and the full variant "ppl-full" or
> "ppl-max", that would Provide "ppl", waiting for the next update of
> gccXY to depend on the alternative (so that versioned depends can be
> used if necessary).
>
> I don't think it is reasonable to aim for further varianting : users who
> are interested in ppl per se, rather than just as a step to build gcc,
> can bear the cost of building one or other interface that interests them
> less..
My immediate goal was to get it working for 10.4 so cloog and gccXY could
be enabled. I don't feel the most qualified to take up ppl, so
contributions are welcome, as are offers to take over. (I'd be glad to
pass it on, not at all offended.) And of course, the offer is extended to
the newer ppl9 (0.11).
fangism
> Jean-Francois Mertens
>
> PS : The "full" variant is for users who want ppl per se; I like to
> build it
> with the latest gccXY, but this is an independent issue.
>
> ------------------------------------------------------------------------------
> Beautiful is writing same markup. Internet Explorer 9 supports
> standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 & L3.
> Spend less time writing and rewriting code and more time creating great
> experiences on the web. Be a part of the beta today
> http://p.sf.net/sfu/msIE9-sfdev2dev
> _______________________________________________
> Fink-devel mailing list
> Fink-devel@lists.sourceforge.net
> http://news.gmane.org/gmane.os.apple.fink.devel
> Subscription management:
> https://lists.sourceforge.net/lists/listinfo/fink-devel
>
David Fang
http://www.csl.cornell.edu/~fang/
http://www.achronix.com/
------------------------------------------------------------------------------
Beautiful is writing same markup. Internet Explorer 9 supports
standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 & L3.
Spend less time writing and rewriting code and more time creating great
experiences on the web. Be a part of the beta today
http://p.sf.net/sfu/msIE9-sfdev2dev
_______________________________________________
Fink-devel mailing list
Fink-devel@lists.sourceforge.net
http://news.gmane.org/gmane.os.apple.fink.devel
Subscription management:
https://lists.sourceforge.net/lists/listinfo/fink-devel
