Closed corneliusroemer closed 1 year ago
Pinning to iqtree >=2
seems to fix it, but only seems to work when I remove pangolin
and pangolearn
- which we probably should not use here due to their complicated dependencies.
Users may need to install these in their separate environment. We use Nextclade in place of pangolin/pangolearn etc, so it makes sense to remove from standard environment.
Hmm. At the beginning of this year we explicitly added pangolin and pangolearn to our Docker runtime (and our ambient runtime install instructions). This is why they're also in our newer Conda runtime here (defined by the nextstrain-base
package). I think a proposal to remove them needs more consideration than was given above.
We use Nextclade in place of pangolin/pangolearn etc
Is that a change from earlier this year?
…which we probably should not use here due to their complicated dependencies.
Their deps are pretty unfriendly to coexistence, but I'm not sure it means we shoudn't use them.
IIRC, they were added before Nextclade could annotate pango lineages.
Back then, we thought it may be worth running Usher ourselves so we have up to date lineage annotations (there were some GISAID delays, and GISAID used pangoLEARN).
In the end, we never ran pangolin (Usher or pangoLEARN) in ncov-ingest.
Happy to hear feedback, whether anyone actually uses pangolin/pangoLEARN from our docker images/conda environment.
Not including Usher makes it not very practical though.
Our ncov workflow still nominally supports running Pangolin when the run_pangolin
config flag is enabled. So I think the questions are:
I found a few uses using GitHub Code Search:
As well as at least one indication of "dark" usage (e.g. usage we can't directly find):
Current Behavior
Currently both osx and linux lockfiles install IQtree v1.6.
One way we could end up with lower versions is if a package moved from conda-forge to bioconda. Then we're stuck with whatever the last version was that was in conda-forge.