#55991 closed enhancement (fixed)
esmf: Add clang variants
| Reported by: | platipodium (Carsten Lemmen) | Owned by: | tenomoto (Takeshi Enomoto) |
|---|---|---|---|
| Priority: | Normal | Milestone: | |
| Component: | ports | Version: | |
| Keywords: | Cc: | ||
| Port: | esmf |
Description
The following seems wrong to me
port install clang-5.0 openmpi-clang50 port clean esmf port install esmf +clang50 +openmpi ---> Computing dependencies for esmf The following dependencies will be installed: openmpi-gcc7
Why does esmf have gcc7 as dependency when I request clang50?
Change History (11)
comment:1 Changed 7 years ago by ryandesign (Ryan Carsten Schmidt)
| Owner: | set to tenomoto |
|---|---|
| Status: | new → assigned |
| Summary: | Mistakenly (?) install openmpi-gcc7 with requested variants +openmpi +clang50 → esmf: Mistakenly (?) install openmpi-gcc7 with requested variants +openmpi +clang50 |
comment:2 Changed 7 years ago by platipodium (Carsten Lemmen)
Yes you're right. My fault (and messing with local port file)
Nonetheless, missing support for clang is an issue. Please change
defect -> enhancement title -> Add support for clang
comment:3 Changed 7 years ago by ryandesign (Ryan Carsten Schmidt)
| Summary: | esmf: Mistakenly (?) install openmpi-gcc7 with requested variants +openmpi +clang50 → esmf: Add clang variants |
|---|---|
| Type: | defect → enhancement |
The port deliberately disables the clang variants, but it's been that way since the port was added, so I don't know why it does that.
comment:4 Changed 7 years ago by platipodium (Carsten Lemmen)
I have been successfully running esmf with clang and openmpi (source install, not macports) for several years, so should not be a problem.
comment:5 Changed 7 years ago by platipodium (Carsten Lemmen)
Variants are inherited from the netcdf-fortran port, which does not have a clang variant. So this needs to be fixed in the dependent port first.
comment:7 Changed 7 years ago by ryandesign (Ryan Carsten Schmidt)
Can you explain why offering these variants here would be desired? Ports usually do not offer variants for choosing a version of clang; MacPorts base chooses for you.
comment:8 Changed 7 years ago by platipodium (Carsten Lemmen)
There is no need for a specific version of clang. I am fine with ports choosing this for me. But currently, clang is expressedly *excluded*. This is a defect, as there is a designated version of ESMF (denoted gfortranclang) that compiles the C++ part of ESMF with clang and the fortran part with gfortran. This is currently not possible with mac ports.
comment:9 Changed 7 years ago by ryandesign (Ryan Carsten Schmidt)
I think that's what the gfortran variant is supposed to do. But this port disables that too, and I don't know why it does that either.
comment:10 Changed 7 years ago by tenomoto (Takeshi Enomoto)
| Resolution: | → fixed |
|---|---|
| Status: | assigned → closed |
comment:11 Changed 7 years ago by platipodium (Carsten Lemmen)
Thanks for the updated to ESMF 7.1.0r that was long-awaited (upstream).
Indeed, with variant gfortran, the cxx compiler used with openmpi is now used by ESMF despite wrong naming of ESMF_COMPILER variable
ESMF_CXXCOMPILER=/opt/local/bin/mpicxx-openmpi-mp ESMF_COMPILER: gfortran /opt/local/bin/mpicxx-openmpi-mp --version
So while this de facto address my request, it is still very confusing for the user (seeing ESMF_COMPILER=gfortran when she expects ESMF_COMPILER=gfortranclang). I don't know of adverse effects upstream, but there might be some ...
I would like to see a changed variant nomenclature (separately choosing one of gcc/clang as c compiler and one of gfortran/g95 as f compiler) and a correct representation in the upstream variable ESMF_COMPILER
When I run
port variants esmfI do not see a variant called "clang50".