The issue has been fixed upstream and will be part of next pip release.

That's been done for years now.

There is no longer any debian branches on the swh-graph repository, so this must have been done already.

 swh  tony  yavin4  ~  work  …  swh  swh-environment  swh-graph   master  1⚑  ERROR  %  git pull
Already up to date.
 swh  tony  yavin4  ~  work  …  swh  swh-environment  swh-graph   master  1⚑  %  git branch --remote | grep -i debian
 swh  tony  yavin4  ~  work  …  swh  swh-environment  swh-graph   master  1⚑  ERROR  %  git branch --remote | grep -i upstream
 swh  tony  yavin4  ~  work  …  swh  swh-environment  swh-graph   master  1⚑  ERROR  %  git branch --remote | grep -i pristine-tar

We should also build the whole documentation when submitting a diff to swh-docs, some sphinx errors not related to packages documentation could be caught this way (see https://jenkins.softwareheritage.org/job/DDOC/job/dev/5700/artifact/swh-docs/docs/errors.log)

Solution 3 has been implemented through new tox environments in each swh package.

So either the vault is fine and the test needs to be improved.
Or the vault builds the tarball without taking into account the symlink case.

@rendong951 Sorry, but we can't reproduce the issue on our end.