Add HDF5 (pt. 2) - merge when passing #199
Conversation
|
cc @ocefpaf Maybe it won't be that hacky at all. It turns out there is a |
| ./configure --prefix=$PREFIX \ | ||
| --enable-linux-lfs --with-zlib \ | ||
| --with-pthread=yes --enable-cxx --with-default-plugindir=$PREFIX/lib/hdf5/plugin \ | ||
| --disable-silent-rules # To make Travis happy with the level of activity. |
There was a problem hiding this comment.
Here's the argument to add a bit more noise to the Travis CI build. Hopefully, that is the right amount without going overboard the other way.
|
Appears this is not an issue with the amount of noise, but actually may be a memory issue with the |
jakirkham
force-pushed
the
add_hdf5
branch
2 times, most recently
from
b7993cf to
5e097ff
Compare
This test appears to be pretty resource intensive and is grinding Travis CI to a halt. As a result, we will simply disable running this test and see if that is sufficient in order to run the rest of the test suite.
| @@ -2,7 +2,8 @@ | |||
|
|
|||
| ./configure --prefix=$PREFIX \ | |||
| --enable-linux-lfs --with-zlib \ | |||
| --with-pthread=yes --enable-cxx --with-default-plugindir=$PREFIX/lib/hdf5/plugin | |||
| --with-pthread=yes --enable-cxx --with-default-plugindir=$PREFIX/lib/hdf5/plugin \ | |||
| --disable-silent-rules # To make Travis happy with the level of activity. | |||
There was a problem hiding this comment.
This attempts to increase the noise level.
|
So, I have added a patch here, which simply does not run the cache program from This appears to allow for the Mac Travis CI build to complete. Without it, the build hangs and is terminated after no response for ~10mins. It is unlikely this patch would be accepted upstream as it skips one of their tests. So, I don't plan to submit it. However, a discussion could begin about making this test less resource intensive (assumed to be memory intensive). Depending on the nature of the test, this may or may not be feasible. |
|
🎉 Travis CI passes! 🎉 |
| url: http://www.hdfgroup.org/ftp/HDF5/current/src/hdf5-{{version}}.tar.gz | ||
| md5: b8ed9a36ae142317f88b0c7ef4b9c618 | ||
| patches: | ||
| - test_Makefile.in.patch |
There was a problem hiding this comment.
I'd want a really good comment in here to explain the rationale. Maybe provide a link back to the original PR too.
P.S. I'm happy for this to be done in the feedstock once merged.
There was a problem hiding this comment.
No problem. Are you ok with the rationale?
There was a problem hiding this comment.
Yep. Slightly concerned - maybe we should have an environment variable which allows us to turn stuff on if we are conda-forge building stuff. Like:
patches:
- test_Makefile.in.patch # [os.environ.get(conda-forge)]
This continues off @groutr's excellent work in this PR ( #192 ) to get a build of HDF5 going.
Initially thought Travis CI was failing builds due to a lack of output from the tests. However, it has been determine that one of the test portions was causing a hang. Namely the
cachetests. Here we simply patch thetest/Makefile.into skip thecachetest program. This appears to be enough to get Travis CI to run through. We go ahead and apply this patch everywhere even though Travis CI was the only one that needed. My best guess is this test used too much memory, which led to hang.