mirror of
https://github.com/facebook/sapling.git
synced 2024-10-10 08:47:12 +03:00
103b3116d4
Summary: Fetching all history of both filenodes to see if there's common history either side of a merge is wasteful, and in some megarepo work is causing long delays deriving merge changesets. Where we have already derived filenodes for a given merge's ancestors, we can go faster; we can use the linknodes to determine the older of the two filenodes, and fetch only history for the newer of the two. This is imperfect for the import use case, since filenodes depend on hgchangesets, and the batching in use at the moment prefers to generate all derived data of a given type before moving onto another type, but it's an improvement for cases where some filenodes are already derived (e.g. due to import of a repo with a similar history). Reviewed By: StanislavGlebik Differential Revision: D28796253 fbshipit-source-id: 5384b5d2841844794a518c321dbf995891374d3a |
||
---|---|---|
.. | ||
blobrepo_hg | ||
blobsync | ||
changeset_fetcher | ||
common | ||
errors | ||
override | ||
repo_blobstore | ||
src | ||
test | ||
Cargo.toml |