(What a nice error message though!)

The status quo for both Biopython and Augur has been to support comments in FASTA files. With version 1.85, Biopython is switching things up. Our options are:

1. Let Biopython decide when Augur should stop supporting comments in FASTA files.
   • Expose this warning, or at least some variant, to our users when reading/parsing FASTA files.
   • The case of parsing with format='fasta' on a GenBank file in ancestral.py is tricky since it appears when attempting to parse a GenBank file as FASTA, however suppression of the warning will also suppress it for when the root sequence is a FASTA with comments.
2. Decide independently if/when Augur should stop supporting comments in FASTA files.
   • This warning should not be exposed to our users when reading/parsing sequence files. It can be suppressed by something like warnings.filterwarnings("ignore", category=BiopythonDeprecationWarning, message="Previously, the FASTA parser …").
   • To continue supporting comments from FASTA files, use format='fasta-pearson' and pin Biopython >=1.85 as that's the first version to support the format. We may not want to add this pin immediately to allow users to install Augur with older versions of Biopython for some more time.

I think (2) is the way to go.

• To continue supporting comments from FASTA files, use format='fasta-pearson' and pin Biopython >=1.85 as that's the first version to support the format. We may not want to add this pin immediately to allow users to install Augur with older versions of Biopython for some more time.

Or don't pin but use feature/version detection to use format='fasta' for <1.85 and format='fasta-pearson' for >=1.85? They're functionally equivalent, right? (Or close to it?)

Here's the full deprecation announcement (unfortunately hidden in DEPRECATED.rst and not at all mentioned in the release notes): https://github.com/biopython/biopython/blob/e8d27e912ee8c9e75483728f6ac8094713563c1d/DEPRECATED.rst#bioseqiofastaio

Or don't pin but use feature/version detection to use format='fasta' for <1.85 and format='fasta-pearson' for >=1.85? They're functionally equivalent, right? (Or close to it?)

I agree this is the way to go. We have no reason to change.

fasta-pearson is unfortunately only supported in >=1.85 (it's missing here https://biopython.org/docs/1.84/api/Bio.SeqIO.FastaIO.html) so we need to do a version switch here, but that's better than having inconsistent behaviour depending on a dependency's version.

Thanks for the reviews @corneliusroemer. I'll try to address things properly in #1730 and #1731, which means bigger changes in code rather than a quick fix. In the meantime, can we disable the failing test to improve usefulness of CI?