Closed Rowern closed 4 years ago
Tree is better for calculating cache IDs. Currently stages inherit cache ID from last step of previous stage - which is not necessary if no files are copied between them.
Also in theory it's possible to execute some stages without "RUN" in parallel, or ignore those that not parent of target.
Maybe this Closes https://github.com/uber/makisu/issues/262
But I saw https://github.com/uber/makisu/issues/280 and don't really know why makisu should implement stages as tree ?