WIP: Update to 4.3.0 and update pinning

[kmuehlbauer]

successor of #7 and #8

[conda-forge-linter]

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe) and found it was in an excellent condition.

[kmuehlbauer]

@ocefpaf

I had to create this new pr, because we need to add something to build.sh also. I have the cmake build tested (see comments), this doesn't work yet, because ncxx4-config isn't generated. So I had to put in the whole stuff for automake and friends.

[pelson]

Looks sensible to me. 👍

[kmuehlbauer]

Strange behaviour here experiencing. If using autoreconf it complains about missing perl.

[kmuehlbauer]

OK, we'll use the precompiled configure until the autoreconf stuff is sorted out. See also https://circleci.com/gh/conda-forge/xraylib-feedstock/1 which fails with the same error. @ericdill any thoughts?

[kmuehlbauer]

@ocefpaf @pelson

I'm not sure, what's going on with the travis builds.

checking build system type... i386-apple-darwin13.4.0

is detected instead of

checking build system type... x86_64-apple-darwin13.4.0

Finally failing with:

checking for gcc... clang
checking whether the C compiler works... no
configure: error: in `/Users/travis/miniconda3/conda-bld/work/netcdf-cxx4-4.3.0':
configure: error: C compiler cannot create executables

Unfortunately config.log can't be displayed after failing.

[kmuehlbauer]

I've hacked .travis.yml to get config.log displayed.

[ocefpaf]

Strange behaviour here experiencing. If using autoreconf it complains about missing perl.

I've seen that before. No idea what is going on 😒

[ocefpaf]

I am guessing the new docker image might be lacking perl. So we need to rebuild autoreconf and add perl as a dependency there. (This might fix a lot of failures I am seeing.)

Pinging @pelson who might sniff if we are in a red-herring search or if this sounds reasonable.

[kmuehlbauer]

Here is evidence, I put a which perl in the build-script:

https://circleci.com/gh/conda-forge/netcdf-cxx4-feedstock/25

+ which perl
which: no perl in (/opt/conda/envs/_build/bin:/opt/conda/bin:/opt/rh/devtoolset-2/root/usr/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin)
Command failed: /bin/bash -x -e /recipe_root/build.sh

[pelson]

Sounds plausible if autoreconf needs perl. I'm afraid I don't have the cycles today to look at this though. Are you happy to take a look at fixing?

[ocefpaf]

Sounds plausible if autoreconf needs perl. I'm afraid I don't have the cycles today to look at this though. Are you happy to take a look at fixing?

Already on it conda-forge/autoconf-feedstock#8

(Save your cycles for better things. I am happy to do this while I watch 6 hours straight of presentations about Data Management 😉)

[kmuehlbauer]

For the travis-failures:

https://travis-ci.org/conda-forge/netcdf-cxx4-feedstock/builds/134728607#L599

configure:3684: checking for gcc
configure:3711: result: clang
configure:3940: checking for C compiler version
configure:3949: clang --version >&5
sh: line 1:   927 Trace/BPT trap: 5       /Applications/Xcode.app/Contents/Developer/usr/bin/xcodebuild -sdk / -find clang 2> /dev/null
clang: error: unable to find utility "clang", not a developer tool or in PATH

[kmuehlbauer]

OK, here is what I have:

1. configure - based build breaks on osx for an unknown reason
2. cmake - based build works, but misses ncxx4-config (see ncxx4-config missing in cmake-build Unidata/netcdf-cxx4#36)

I'll stop iterating here.

We could hack something together to get ncxx4-config working, but it would be better to have it right from the sources. ping @WardF

[WardF]

I'm at a workshop today and tomorrow but I'll address the ncxx4-config issue that's been opened. I wonder if the problem with configure on OSX is related to the issue discussed here: conda-forge/staged-recipes#213 (comment)

[kmuehlbauer]

@WardF Thanks for your quick response, it is very welcome. Thanks for the osx-pointer, I'll have a look into this.

[kmuehlbauer]

I'll clean up everything, when the cmake-ncxx4 issue is sorted out, also if this means to wait some time.

I have the hardest time to get autoconf/automake running properly, so switching to the working cmake is probably best.

[ocefpaf]

I have the hardest time to get autoconf/automake running properly, so switching to the working cmake is probably best.

👍 (Although I am trying cmake with netcdf-fortran and failing right 😁)

[jakirkham]

So what was the issue with autoconf(?) and perl?

[jakirkham]

NVM I see it. Makes sense. Good detective work @kmuehlbauer. (I know it was the closest thing I could find 😉)

[kmuehlbauer]

I have PR Unidata/netcdf-cxx4#37 ready, which fixes the ncxx4-config-issue with the cmake builds. Hopefully the guys at unidata find some time to get this in.

[kmuehlbauer]

Since at unidata everything is green in their build, I have added this as a patch here to move forward.

[ocefpaf]

Since at unidata everything is green in their build, I have added this as a patch here to move forward.

Agreed. That is a fine strategy to backport things. (I am still stuck with netcdf-fortran, but I will try again tomorrow.)

[WardF]

I'll review the pull request today and get it merged unless something unusual occurs!

[kmuehlbauer]

@WardF It seems, that there is some glitch within the build ncxx4-config left, which I could not figure out yet.

[WardF]

No problem; I really appreciate the pull request as it's a great starting place. I've got some other stuff on my plate (Unidata/netcdf-c#239) but will turn my attention to this as soon as possible, I don't think it will take much to get it the rest of the way. Thanks again!

[ocefpaf]

@kmuehlbauer should we merge this?

[kmuehlbauer]

@ocefpaf The ncxx4-config is broken or is reporting wrong setup.

But it should work, I'm 👍 for merge.

[ocefpaf]

@ocefpaf The ncxx4-config is broken or is reporting wrong setup.

Maybe we should remove the patch that adds it and leave ncxx4-config out, like what we did for netcdf-fortran, no? (I can do that BTW if you are busy.)

[kmuehlbauer]

@ocefpaf Sounds very reasonable, and yes if you could overtake 👍

[ocefpaf]

Something changed (maybe in the docker image?) and now the Linux builds are broken 😒

[ocefpaf]

Something changed (maybe in the docker image?) and now the Linux builds are broken 😒

It seems to be the toolchain 😒
See dropping it )9d261e9) solved the issue in #10

[kmuehlbauer]

@ocefpaf

The ncxx4-issue is solved at unidata (Unidata/netcdf-cxx4#37). How do we proceed. Bump build number?

[ocefpaf]

The ncxx4-issue is solved at unidata (Unidata/netcdf-cxx4#37). How do we proceed. Bump build number?

The changes are not in a release, right? If so we need to backport the fix via a patch and bump the build number. But if release 4.3.1 is near we can just upadte to that once it is out.