Details
-
Suggestion
-
Resolution: Fixed
-
None
Description
Currently we use full clones during:
- collecting changes, on bamboo server
- code retrieval, on agents, during performing build
In both cases we don't need the whole history, so if it is possible we should do shallow clones.
Shallow clones are not implemented in JGit yet.
Also - both server and agents should be able to fetch afterwards properly. Corner test case:
- detecting new rev A -> add to queue
- detecting new rev B (grandchild of A) -> add to queue (concurrent builds are on)
- agent -> takes rev B, does shallow clone on it (so it won't have rev A)
- agent -> takes rev A, does fetch - in such case agent should somehow obtain rev A
Attachments
Issue Links
- details
-
BAM-7452 Investigate whether shallow clones (Git Repo) are fast enough
- Closed
-
BAM-7655 Git Plugin should detect if remote repository support 'shallow' protocol
- Closed
-
BAM-7674 Git repo shallow clones should work via http(s)
- Closed
-
BAM-7788 Git repo initial cache retrieval (the shallow one) should work properly between multiple plans using the same repo.
- Closed
-
BAM-7789 (Git repo) Successive code retrievals should use non-shallow fetches
- Closed