NEW: minpack

[jjstickel (Jonathan Stickel)]

New port for minpack, the legacy fortran libraries for solving nonlinear equations and nonlinear least squares problems.

Portfile will be attached. It is not a "normal" install, and so the Portfile could use some review and possible revision. I followed some of the suggestions here: ​http://devernay.free.fr/hacks/netlibs.html

[jjstickel (Jonathan Stickel)]

minor revision

[jmroot (Joshua Root)]

You don't need to handle the patchfile yourself just because it's remote and gzipped. Patchfiles that are not present in files/ will be fetched and applied automatically. They can also be gzipped.

Putting --prefix=${prefix} in configure.args is redundant, as it already appears in configure.pre_args.

[jjstickel (Jonathan Stickel)]

Replying to jmr@…:

You don't need to handle the patchfile yourself just because it's remote and gzipped. Patchfiles that are not present in files/ will be fetched and applied automatically. They can also be gzipped.

This is helpful, but I am not sure how to implement it. Can you show me an example? Will the patchfile just be a url?

[jjstickel (Jonathan Stickel)]

I revised how the patchfile is applied and removed configure.args, as suggested. I think the latest attached Portfile is ready.

[mf2k (Frank Schima)]

A few questions / issues:

1. Isn't this already in py-scipy?
2. I don't like the version handling. You hardcode it many times in the portfile. Since it is not used directly, why not use version "19961126" instead of "19961126.16" and call "16" the minor_version or similar and then use those variables throughout the portfile?
3. Do we really need gcc43 and gcc44 variants? Would anyone actually use that in a new port?

[mf2k (Frank Schima)]

Furthermore, you might as well add a gcc48 variant.

[jjstickel (Jonathan Stickel)]

Replying to macsforever2000@…:

A few questions / issues:

1. Isn't this already in py-scipy?

Yes, but compiled using f2py (I think) specifically for use in python. I am creating this port so that the library can be linked from fortran programs (and I suppose c/c++, but I haven't tested).

1. I don't like the version handling. You hardcode it many times in the portfile. Since it is not used directly, why not use version "19961126" instead of "19961126.16" and call "16" the minor_version or similar and then use those variables throughout the portfile?

OK, I can do this. This is legacy code, and hence version numbers will only change if patches are necessary for compiling with modern compilers, but nonetheless I understand your request.

1. Do we really need gcc43 and gcc44 variants? Would anyone actually use that in a new port?

I can certainly remove gcc43 and gcc44 and add gcc48. I had copied these variants over from some other port.

Thank you for your input.

[jjstickel (Jonathan Stickel)]

How about the latest Portfile that I have attached?

[mf2k (Frank Schima)]

Added in r106900 with gcc47-49.