I manually applied the Python 3.9 one and that passes, so it doesn't seem to be solvability iusse: conda-forge/mlprodict-feedstock#1
it is not in the graph: https://github.com/regro/cf-graph-countyfair/blob/master/node_attrs/mlprodict.json
this gives an error
I found the issue
someone is making tags for each branch in their feedstock and it is confusing git
caused the list of feedstocks to be out of date
The bot is failing (silently) on one of my repos. It used to work just fine every time the PyPI package was updated, but lately it's been missing updates intermittently. On the status page, under "Errored", you can see the quaternion package, and it just says
No error information available for version '2021.3.17.16.51.43'. Any ideas?
1 reply
2021-03-18T00:01:31.8118243Z INFO:conda_build.variants:Adding in variants from /home/runner/work/autotick-bot/autotick-bot/cf-graph/feedstocks/quaternion-feedstock/recipe/conda_build_config.yaml
2021-03-18T00:01:31.8138109Z Traceback (most recent call last):
2021-03-18T00:01:31.8141957Z File "/usr/share/miniconda/envs/run_env/bin/conda-smithy", line 10, in <module>
2021-03-18T00:01:31.8143135Z sys.exit(main())
2021-03-18T00:01:31.8145087Z File "/usr/share/miniconda/envs/run_env/lib/python3.9/site-packages/conda_smithy/cli.py", line 613, in main
2021-03-18T00:01:31.8148049Z args.subcommand_func(args)
2021-03-18T00:01:31.8149261Z File "/usr/share/miniconda/envs/run_env/lib/python3.9/site-packages/conda_smithy/cli.py", line 419, in __call__
2021-03-18T00:01:31.8152134Z self._call(args, tmpdir)
2021-03-18T00:01:31.8154100Z File "/usr/share/miniconda/envs/run_env/lib/python3.9/site-packages/conda_smithy/cli.py", line 424, in _call
2021-03-18T00:01:31.8154923Z configure_feedstock.main(
2021-03-18T00:01:31.8156049Z File "/usr/share/miniconda/envs/run_env/lib/python3.9/site-packages/conda_smithy/configure_feedstock.py", line 1987, in main
2021-03-18T00:01:31.8160212Z render_azure(env, config, forge_dir, return_metadata=True)
2021-03-18T00:01:31.8161822Z File "/usr/share/miniconda/envs/run_env/lib/python3.9/site-packages/conda_smithy/configure_feedstock.py", line 1222, in render_azure
2021-03-18T00:01:31.8163188Z return _render_ci_provider(
2021-03-18T00:01:31.8165243Z File "/usr/share/miniconda/envs/run_env/lib/python3.9/site-packages/conda_smithy/configure_feedstock.py", line 602, in _render_ci_provider
2021-03-18T00:01:31.8166309Z ) = conda_build.variants.get_package_combined_spec(
2021-03-18T00:01:31.8167573Z File "/usr/share/miniconda/envs/run_env/lib/python3.9/site-packages/conda_build/variants.py", line 596, in get_package_combined_spec
2021-03-18T00:01:31.8172209Z combined_spec = combine_specs(specs, log_output=config.verbose)
2021-03-18T00:01:31.8173492Z File "/usr/share/miniconda/envs/run_env/lib/python3.9/site-packages/conda_build/variants.py", line 322, in combine_specs
2021-03-18T00:01:31.8174764Z values = _combine_spec_dictionaries(specs, extend_keys=extend_keys, zip_keys=zip_keys,
2021-03-18T00:01:31.8176193Z File "/usr/share/miniconda/envs/run_env/lib/python3.9/site-packages/conda_build/variants.py", line 294, in _combine_spec_dictionaries
2021-03-18T00:01:31.8179281Z raise ValueError("variant config in {} is ambiguous because it\n"
2021-03-18T00:01:31.8180927Z ValueError: variant config in /home/runner/work/autotick-bot/autotick-bot/cf-graph/feedstocks/quaternion-feedstock/recipe/conda_build_config.yaml is ambiguous because it
2021-03-18T00:01:31.8182254Z does not fully implement all zipped keys, or specifies a
2021-03-18T00:01:31.8183008Z subspace that is not fully implemented. To be clear:
2021-03-18T00:01:31.8183892Z .. we did not find ['1.16'] from {'numpy': ['1.16']} in numpy:['1.17', '1.17', '1.17']
2021-03-18T00:01:31.9498223Z Adding in variants from internal_defaults
2021-03-18T00:01:31.9499580Z Adding in variants from /tmp/tmpqus470ht/conda_build_config.yaml
2021-03-18T00:01:31.9502763Z Adding in variants from /home/runner/work/autotick-bot/autotick-bot/cf-graph/feedstocks/quaternion-feedstock/recipe/conda_build_config.yaml
best to remove the numpy pinning
getting the bot to output a better error is def needed!
Maybe this is the wrong place to ask, but I do actually need numpy as part of the build (for the C-API), so I thought it had to be pinned. Is that no longer the case?
the issue here is that your pin is conflicting with our global ones
so if you remove that file and rerender, then you'll get global pins
they are generally fine unless your package has special requirements
so i'd just use them
Seems like I should document that once I know how it works.
There might be a better way I normally use
ripgrep with a local clone of https://github.com/regro/libcfgraph
we can run it over all packages and make the recipes "searchable"
(but the github search in the libcfgraph repo might also work fine for the moment)
agreed a searchable thing online would be much better
I keep a notebook that loads the graph lying around
can be useful for such tasks
We're trying to figure out if there is a way to get the autobot to do prereleases or releases on branches other than the main branch of a feedstock (for example, previous release branches that are still in maintenance). Nick found something that seemed maybe relevant: conda-forge/jupyterlab-feedstock#267
(I think there is an issue somewhere for it, I'll look it up if you're interested)
This is the failing URL, https://github.com/cctbx/cctbx_project/releases/download/v2021.3/cctbx-2021.3.tar.gz
I released on PyPI a version 0.5.0rc2 of my
docs-versions-menu package which has an existing feedstock. So far, regro-cf-auotick-bot hasn't opened a PR on the feedstock for the new release. Is this expected? Some of the comments here seem to indicate that the bot doesn't react to pre-release version. Is there anything I can do to make the bot generate the new release of the conda package, or will have I have make my own PR by hand?