search

dandi / dandisets

730 Dandisets, 807.1 TB total. DataLad super-dataset of all Dandisets from https://github.com/dandisets
10 stars 0 forks source link

Ensure latest version of `backups2datalad` is installed on every run #381

Closed jwodder closed 7 months ago

jwodder commented 7 months ago

The backups are currently failing to back up embargoed Dandisets because the copy of backups2datalad in the Conda environment was not upgraded after https://github.com/dandi/backups2datalad/pull/12 was merged. This is because, when installing from a VCS URL, pip install (with or without --upgrade) will only update an already-installed package if its version number has increased (regardless of whether any changes have been made to the code), yet backups2datalad's version is constantly at 0.0.0 due to our lack of a need for versioning.

This PR solves the problem by uninstalling backups2datalad before running pip install, so that the latter will have no choice but to install the latest revision.

yarikoptic commented 7 months ago

yikes, that's an odd behavior of pip there IMHO. Ok, let's try this dance. Thank you @jwodder for troubleshooting it