https://pantsbuild.org/ logo
Join Slack
Powered by
Where do we compute the fingerprint of a dataclass...
# development
h
Where do we compute the fingerprint of a dataclass (e.g., a Process)?
h
I believe the hash function, which is auto-done by dataclass. For some intrinsics, we now define them in Rust and manually implement hash. See fs.rs
h
It's only guaranteed to return the same result for the same input within a single process
I hope not, because Python hash() isn't consistent across processes
But I doubt that's what we're using since we'd get no cache hits across a pantsd restart, and presumably wed have noticed
f
for CAS purposes, the cache key is the sha-256 hash of the encoded REAPI protobuf
👍 1
at least for remote execution
h
Where in the code is this computed?
I'm trying to debug a cache miss that shouldn't be
f
the action digest is being used as the cache key
(the REAPI action includes the command digest and input root digest)
do you have debug logs of the command/action protos? if not, maybe add a debug log of the action and command protos in the `digest` function and see exactly what is being computed
`-ldebug`does print the protos out so would be helpful to see logs from both runs
h
Cool, that is very helpful
Open in Slack
PreviousNext
Powered by