~sci-libs/netcdf-4.1.1 builds an unusable fortran library, all the functions declared there appear to be defined with one underscore too many. A part of netcdf-4.0.1-cfortran.patch is still needed to solve this.
Created attachment 257388 [details, diff]
The part of the old netcdf-4.0.1-cfortran.patch still required by netcdf-4.1.1.
Created attachment 257389 [details, diff]
A patch to the sci-libs/netcdf-4.1.1-r3.ebuild to include the netcdf-4.1.1-cfortran.patch.
+*netcdf-4.1.1-r4 (17 Dec 2010)
+ 17 Dec 2010; Kacper Kowalik <firstname.lastname@example.org> +netcdf-4.1.1-r4.ebuild,
+ Default to gfortran if cannot figure out compiler wrt #348922 by Honza
+ Macháček <Hloupy.Honza@centrum.cz> Drop old.
Unfortunately, that's only workaround. We are defaulting to what we think is sane value. Though now netcdf will fail with mpi wrapped around intel compiler. Leaving bug open until fixed properly.
What compiler do we want to support? The current netcdf detection is limited to a very small set of compilers. The user is expected to set a CPPFLAGS appropriate to their fortran compiler in most case. I learned that the hard way on a BlueGene/P using the IBM compiler.
We could extend the detection in Gentoo but we need to be clear about which compilers we will support.
+ 28 Feb 2015; Justin Lecher <email@example.com> -netcdf-4.1.1-r4.ebuild,
+ -netcdf-4.1.3.ebuild, -netcdf-220.127.116.11.ebuild, -netcdf-4.3.0.ebuild,
+ -netcdf-18.104.22.168.ebuild, -netcdf-4.3.2.ebuild, metadata.xml:
+ Drop old; obsoletes #348922, #363131, #435844, #454950 & #478618