Pin to hdf5 1.8.17*

[ocefpaf]

@kmuehlbauer I am pinning against the latest hdf5 available in conda-forge. This will be used in all packages soon. Please take a look at it and merge if you think it is OK.

Needs: conda-forge/netcdf-cxx4-feedstock#7 and conda-forge/libnetcdf-feedstock#1

[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 We need to get netcdf-cxx4 ready first.

[kmuehlbauer]

@ocefpaf @jakirkham The problem with hdf5 was that conda wasn't able to pull in the correct hdf5 version at runtime.

So the dependency chain is as follows:

libnetcdf --> netcdf-cxx4 --> libradolan

So the only place where hdf5 comes into play is libnetcdf. If all resolves properly, we do not need to hdf5-pin any more.

[ocefpaf]

Cool! I will update this PR once we get the others merged to see if we can drop the dependency over-specification.

[kmuehlbauer]

@ocefpaf This is blocked atm by netcdf-cxx4, which depends on fixing ncxx4-config at unidata: Unidata/netcdf-cxx4#37

[kmuehlbauer]

@ocefpaf Can you restart the CI's here? Now, that the netcf-cxx4 thing is sorted out, we could try this here again.

[kmuehlbauer]

@ocefpaf nvm I just noticed that I could do it myself

[ocefpaf]

@kmuehlbauer this will need more thought. Does libradolan expects ncxx4-config to exist?

[kmuehlbauer]

@ocefpaf I'll have a look into this. Thanks!

[kmuehlbauer]

@ocefpaf Found it. We're creating wrong library names in https://github.com/conda-forge/netcdf-cxx4-feedstock. It should be libnetcdf_c++4, but we're creating libnetcdf_cxx4. Any clues?

[ocefpaf]

I guess the name changed in the update. I need to check netcdf-cxx4 code to be sure.

[kmuehlbauer]

Unidata is also producing libnetcdf_c++4 (travis-log). Merkwürdig.

[ocefpaf]

I'll take a look. The only changes are the use of cmake and the update.
Maybe we should create a symlink?
Em 21/06/2016 06:50, "Kai Mühlbauer" notifications@github.com escreveu:

Unidata is also producing libnetcdf_c++4 (travis-log
https://s3.amazonaws.com/archive.travis-ci.org/jobs/130138115/log.txt).
Merkwürdig.

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
#3 (comment),
or mute the thread
https://github.com/notifications/unsubscribe/AA6BL4tHSQSd4UdVP5LcEKbp93ubVM8zks5qN7PSgaJpZM4IqFLt
.

[kmuehlbauer]

Unidata is also producing libnetcdf_c++4 (travis-log).

This was only half the truth. Obviously the cmake builds also at Unidata creates libnetcdf_cxx4 whereas the configure build creates libnetcdf_c++4.

@ocefpaf As you already suggested, a symlink should do the trick. Or just renaming as it mimicks the configure behaviour. WDYT?

[ocefpaf]

WDYT?

Lets's symlink and cover both naming. Will do soon. Just trying to find a bug in conda-smithy now...

[ocefpaf]

@kmuehlbauer conda-forge/netcdf-cxx4-feedstock#11 should fix the issue here.

[ocefpaf]

@kmuehlbauer I updated the boost pinning here. Is that OK?

[ocefpaf]

If we pin netcdf-cxx4 we don't need to add the hdf5 pin here b/c the right libnetcdf/hdf5 will be downloaded.

[kmuehlbauer]

If we pin netcdf-cxx4 we don't need to add the hdf5 pin here b/c the right libnetcdf/hdf5 will be downloaded.

That had historic reasons b/c that time this fixed some wrongly pulled hdf5. Now it won't be needed any more.

[kmuehlbauer]

Travis Problems, we have to wait.

[kmuehlbauer]

@ocefpaf We had this before, I think, conda-forge/netcdf-cxx4-feedstock#5 This has to be some problem with libc++ / stdlibc++. We should have cmake and make output more verbos to see what is going on here.

[ocefpaf]

@ocefpaf We had this before, I think, conda-forge/netcdf-cxx4-feedstock#5 This has to be some problem with libc++ / stdlibc++. We should have cmake and make output more verbos to see what is going on here.

There is an ongoing conversation on gdal/hdf5 feedstock about this. I don't have a Mac to follow that, but I guess we should read those threads before trying anything here.

[kmuehlbauer]

There is an ongoing conversation on gdal/hdf5 feedstock about this. I don't have a Mac to follow that, but I guess we should read those threads before trying anything here.

We should also not forget, that netcdf-cxx4 will have the same problems. Hopefully this will be sorted out soon.

[ocefpaf]

@kmuehlbauer success! 🎉

[jakirkham]

Looks clean AFAICT.

[kmuehlbauer]

@ocefpaf @jakirkham great news. Will check this soon. Now let's merge this.