“Stash” is the current working name for a command that can be used to temporarily suspend patches from the active set in a repository and then restore them with the same identity. This in contrast to rebase, which always restores patches with a different identity.

Protoype implementation at

To use it, darcs stash suspend and darcs stash unsuspend.

Any comments to


The implementation is only very lightly tested and is likely to go wrong when the suspended patches get into complicated conflicts.

The name isn’t great as it this isn’t really the same as git stash which captures unrecorded local changes. Maybe suspend and unsuspend should be top-level commands of their own.

In the long-term, we should unify this with rebase, for example making it easy to switch from a normally suspended patch to one that’s had its identity changed.

Should we enable direct stashing of unrecorded changes?

The implementation is some way from being mergeable as it refactors a whole load of darcs code first (including a lot of the rebase implementation).


Suspending patches locally provides a sort of halfway-house to the full idea of multiple branches

There is also the pre-existing workflow of using darcs obliterate -O to shelve patches. Darcs stash can be understood as being more robust wrt dependencies. Suppose you have a patch “Later” which depends on “Earlier”. If you obliterate -O “Later” and then amend/delete “Earlier”, the “Later” patch cannot be applied (not even with darcs rebase apply). With the darcs suspend/unsuspend workflow, darcs will remember to also suspend “Earlier” if at any point you modify or remove it.

Some relevant IRC discussions: