effff869a1
A checkpoint write is unnecessary if it does not change the semantics of the data currently stored in the checkpoint. We currently perform unnecessary checkpoint writes in two cases: - Same steps where no aspect of the resource's state has changed - Replace steps, which exist solely for display purposes The former case is particularly bothersome, as it is rather common to run updates--especially in CI--that consist largely/entirely of these same steps. These changes eliminate the checkpoint writes we perform in these two cases. Some care is needed to ensure that we continue to write the checkpoint in the case of same steps that do represent meaningful changes (e.g. changes to a resource's output properties or dependencies). Fixes #1769. |
||
---|---|---|
.. | ||
apitype | ||
backend | ||
diag | ||
encoding | ||
engine | ||
graph | ||
operations | ||
resource | ||
testing | ||
tokens | ||
tools | ||
util | ||
version | ||
workspace |