-
Notifications
You must be signed in to change notification settings - Fork 90
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Fix default file format when using legacy netcdf #2062
Conversation
@bshanahan Please could you check this works for you? |
I am not sure legacy netcdf is the best name. It also breaks on ubuntu 18.04, failure from
Can we rename so that |
I have theses
|
Maybe something like I do want to eventually remove the legacy version completely -- it's almost ten years old now! Plus maintaining two separate versions is not great. That error is because we've repeated setting the default file format. I'll fix that. Also, just to clarify the different versions that are available, because the naming is a bit of a mess:
Other distros may call the C++ API something like |
Still broken for me. |
Please could you post the error message and/or config.log? I'm afraid I don't recall what the original issue was |
`datadir` and `dump_format` are set in `BoutInitialise`, so must already have values. Setting a default value is either redundant or inconsistent
Sorry, the original problem is that configure did not recognize netcdf, so switched to h5. This has been fixed. I got confused because today I mentioned on slack that my code breaks if I try and collect while a simulation is running. This PR does not fix that problem. |
* next: always pull fedora from fedoras registry Allow to override packages in fedora rawhide Delete removed test_io.grd.nc file from test-invpar's CMakeLists.txt SurfaceIter iterate over non-guard cells by default Formatting changes Invpar fixes for non-uniform dy CI: Turn off patch coverage check Prefer references to pointers Add variation of g_22 metric, dz != 0 Remove commented-out 'grid=' line in input file for test-invpar Replace ccoef=3 test in test-invpar Use non-integer dcoef in test-invpar to avoid zero-pivot error Use FFT z-derivatives for test-invpar Don't use grid file for test-invpar Remove incorrect dz factors in InvertParCR Add z-variation to input of test-invpar to properly test z-derivs Correct implementation of Grad2_par2 in InvertParCR Check location of coefficients in InvertParCR Minor optimization of cyclic InvertPar implementation Change typo in curvature section
Also (hopefully) fixed the problem @dschwoerer reported with |
Removing the redundant default options in physicsmodel affects things that don't set the default options... My comeuppance for not using BoutInitialise!
* next: (61 commits) Fix links to open-mpi.org and lam-mpi.org CMake: Add option to skip generating fieldops CMake: Make generated_fieldops.cxx command more portable CMake: Add module to find clang-format Suggestions for gnu build on Marconi Minor tidying of test-io_hdf5 runtest (including black formatting) Update the benchmark file for the hdf5 version of test-io too Minor tidying of test-io runtest (including black formatting) Update test-io benchmark file for new FieldPerp default attribute Add CMake code to generate generated_fieldops Fix comparison Make sure test-restart-io sets exit code correctly on failure Don't read/write FieldPerps in guard cells Use last boundary point for global y-index in restart IO test Test attributes of written variables in test-io and test-io_hdf5 Better fix for reading/writing FieldPerp index Rename *local -> *global argument in mesh index routines Remove deprecated YLOCAL() in examples/laplace-petsc3d Fix case-sensitivity of __contains__() __eq__() comparison operator for BoutOptions() ...
I checked the setup in #2182 again, and it's skipping lots of tests with
I'd expect the |
`BoutInitialise` ensures these are set, so only an issue if that isn't called
@johnomotani Please can you check now? |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
clang-tidy made some suggestions
Unfortunately not, I still get the same symptoms as my previous comment. Here's the config.log in case that helps. |
* next: (301 commits) GHA: Increase test timeout to 6 minutes Apply black formatting to test-squash Ignore some variables when comparing squashed outputs in test-squash Update expected results of test-io_hdf5 clang-tidy suggestion Fix typo clang-tidy fixes Use bout_type="string" for strings in H5Format Write descriptions for std::vector<int> and std::string variables Fix reading of char* in Ncxx4 Use MPI_C_BOOL type instead of C++ binding MPI::Bool Manual improvements for hypnotoad Removing hypnotoad grid generator clang-tidy suggestion Support location in Mesh::get() Fix a couple of clang-tidy warnings Completely remove LaplaceShoot CMake: Set PYTHONPATH for tests Remove some other references to 3D metrics Convert test-backtrace runtest to Python ...
I found a better way of testing it on my machine -- this should actually work now! |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
clang-tidy made some suggestions
Ah, now it's trying to compile Ncxx4 and OptionNetCDF because |
🤦 I guess this worked for me because the netcdf-cxx4 headers are in a default search path, so the compiler could find them and it worked. I uninstalled it completely and the latest commit should actually, really, work this time! |
There are some tests for OptionsNetcdf that need to be switched off too: |
test-io is expected to fail with the legacy interface.
It does work now.
With a couple more tiny changes, this now works for me. I had to skip |
No description provided.