How would I go about using

adhoc_tool

with a binary that uses its own caching mechanism to avoid redoing work? The cache is both an output and an input to the binary; is there a way to model this? This is if I wanted to use Pants with another build system without rewriting that system in Pants. For a simple example, imagine Pants didn't have Go support, and I had a Python repo using Pants where I wanted to add a Go step that consumes files output by some Pants targets (e.g. protobuf codegen) and outputs files used by other Pants targets (e.g. by a

python_distribution

). If I wanted to do this quickly without writing my own plugin for the language, I'd want to just use the Go

system_binary

, but

go

, like many modern languages, has its own build system with a caching mechanism.

Pants has an internal mechanism for this (“named” or “append-only” caches), but I don’t think it’s exposed to adhoc_tool yet. Could you file a feature request issue?