This is a good point. Ideally, this should require having Nextlow to clone the repo as a bare repository.

Then, when the execution is launched it should check it out into a sub-directory named as the requested versions

I would have also expected main.nf to be read once and kept into RAM (or copied on disk into the execution directory) while nextflow executes the pipeline. Instead, I experience the same issue as above in the following use case:

cd MyGitRepo/
git checkout master
nextflow main.nf
git checkout newFeatureBranch
nextflow main.nf
git checkout otherFeatureBranch

The execution of the sub-processes will execute the state in which the code is at the time where those processes are run, not the state in which they were when the pipeline was launched. I may be wrong but it seems that the only solution is to have one working repository per execution?

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

Solving this Issue would be very useful for our team, as this behavior requires being extra careful when developing and running tests on development pipelines.

Seems to be covered by #2870 which is in our next release milestone