molecule must be able to clone dependencies without overriding them

[ssbarnea]

Issue Type

• Feature request

Molecule and Ansible details

At this moment molecule is not able to clone git dependencies in a way that does not override them if they are already cloned.

gilt approach does always override the destination folder due to https://github.com/metacloud/gilt/issues/65 -- which means that it is not safe to be used on developer machines or on zuul ci which does pre-clone the siblings repositories before running molecule. Initially I wanted to fix gilt itself but I make some discoveries that could prevent a good implementation on it: gilt does not clone the repository to the target location, instead if does copy the files from its own internal clone/cache. This is problematic as it make the decision of shoudl I copy or skip less clear: without a clone at destination what should we do? The discussion should be continued on gilt bug mentioned.

galaxy is not able to clone repositories that have multiple ansible roles because it expects a specific file structure on these repositories. We cannot do anything about galaxy because is deprecated, even new bugs are automatically closed by a bot.

shell seems to be the only method that can temporary be used to workaround this limitation as it can outsource that logic to a shell script. Still, this is quite inconvenient.

mazer - no supprt in molecule for it and in addition to this, it does not seem to support generic repositories either, so is similar to galaxy in this aspect.\

Any ideas?

[decentral1se]

For now, I think shell and a script with git clone is the best we can do here in Molecule ...

Where is the right upstream to get a ticket going? Gilt is cool but ansible/* projects make more sense in terms of future sustainability. Have you done some detective work on #ansible-devel or #ansible-community?

(I know you know but this isn't something we can fix in Molecule in general. Molecule just drives other tools and this fix will have to be in another tool. So, I am not sure we should leave this issue open beyond getting some feedback ... maybe also try the mailing list?)

[decentral1se]

Is ansible/mazer#24 relevant?

[ssbarnea]

Fixed by #2875 and #2877