-
-
Notifications
You must be signed in to change notification settings - Fork 14
Fix building with hdf5 and migrate everything #115
Conversation
…nda-forge-pinning 2022.03.25.20.30.42
PyNIO does not build with the new HDF5 1.12 API, using the API compatibility macros for version 1.10 fixes it.
GCC 10 changes how to handle tentative definitions, with this codebase this generates errors like (simplifying a bit): [... libnio.a(NclNewHDF5.o):(.bss+0x0): multiple definition of `possibleDimNames'; [...] libnio.a(NclHDF5.o):(.bss+0x0): first defined here A workaround is available to use the old behaviour, fixing this issue, see also https://gcc.gnu.org/gcc-10/porting_to.html (par. Default to -fnoc-common). Probably this is not compatible with clang.
MNT: Re-rendered with conda-build 3.21.8, conda-smithy 3.19.0, and conda-forge-pinning 2022.03.25.20.30.42
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 ( I do have some suggestions for making it better though... For recipe:
Documentation on acceptable licenses can be found here. |
Tests are failing because http://test.opendap.org/opendap/data/nc/bears.nc does not work anymore. A few hours ago it was working fine, I hope it's not because of my test runs... |
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 ( |
Try:
|
The server used to download a file needed for this test just went offline. Maybe revist this in the future?
It was added by mistake in commit 45398aa
@conda-forge-admin please rerender |
…nda-forge-pinning 2022.03.25.20.30.42
I think I do not understand how to do migrations correctly, the bot rerendered with older versions. I have not committed migration files to my fork. |
I think it is good to keep the rerender commits completely isolated from your changes. You can then rebase and cleanup your history with more ease |
Are you interested in becoming a maintainer? |
First, thank you for your proposal. I do not know if it is a good idea, I've just improved a bit what I did back in January and sent a similar PR, I do not use any of this stuff and neither understand what it is used for (well, I understand it reads netCDF files, that's all). Back then I was playing with conda for the first time trying to reproduce an environment with newer packages and the first thing that failed was pynio, requiring an old hdf5 version. By luck I found about the compatibility macro, I tested it and for once instead of throwing everything away and filing an issue I decided to make a PR.
I use git in my personal projects and never had to rebase anything, I can try. I do now know if I understand completely what you said, I've made commits with changes and a re-render like 45398aa and 5488686... do you mean that also these commits should be changed? |
I feel like the requirements are quite loose. Ultimately, if you need this even a little bit, it is in your interest to keep it alive. Otherwise, do you use this package at all? I'm somewhat interested in archiving this package since it hasn't been refreshed in a while. But... Maybe we can merge this and this about this problem later. |
A research friend showed me this ugly setup with a hand-made VM with many (many) dependencies, I tried to reproduce it with conda, after trying for 2 days I gave up, but at least I got pynio updated. I think the current solution is to ship back and forth a VM that keeps growing in size (I did not know before how hopeless it was).
Now at least it can be refreshed, maybe having found these fixes will make it easier in the future. |
@conda-forge-admin please rerender |
Hi! This is the friendly automated conda-forge-webservice. I tried to rerender for you, but it looks like there was nothing to do. This message was generated by GitHub actions workflow run https://github.com/conda-forge/pynio-feedstock/actions/runs/2056049142. |
See conda-forge/av-feedstock#53
You can then delete the rerenders when you want to cleanup, or just sqaush them to better understnad what your modifications were |
Link for the file used in this test is now up again, I reverted this change. I also added another patch that fixes a subset of the warnings from numpy, does the build number need to be incremented?. Can new commits still be added to this PR? |
No. please open a new PR. At this stage you are doing the maintainers job. so please add yourself as one. |
"a maintainer's job" |
Checklist
0
(if the version changed)conda-smithy
(Use the phrase@conda-forge-admin, please rerender
in a comment in this PR for automated rerendering)This fixes (at least on linux) #100, #90, maybe it helps with #114.
See commit messages for more info, with these changes pynio builds fine at least with linux_64_numpy1.19python3.9.____cpython and linux_64_numpy1.21python3.10.____cpython
There are 101 passing tests during the build, with 506 warnings.