Should we make

find_binary

uncachable? It relies on the external state on the machine, and things will break if suddenly a certain binary is no longer available at the original path. Or if you added a new binary and want it to be recognized

makes sense to me

would you hash the contents of the binary?

BinaryPaths

only finds the path to the binary, rather than the actual binary. But maybe it should store a private hash! That would properly fix the issue that motivated me to make this change and that I didn’t actually fix Here’s the dataclass:

@dataclass(frozen=True)
class BinaryPaths(EngineAwareReturnType):
    binary_name: str
    paths: Tuple[str, ...]

and with pyenv, all of the shims have the same content

ah, that's troublesome

Well that’s sad. Even tho Pyenv would still be broken, do you think it’s still a good idea to hash the binaries and to follow symlinks? Or maybe only hash the binaries but don’t follow symlinks

you can

pyenv which python

to have it print out the path

I was going to say that generally this can’t be linked to Pyenv, but actually I think we can add that extra logic for our

PexEnvironment

abstraction which finds the interpreters by calling the super generic

BinaryPaths

rule We need a way to identify if it’s a Pyenv interpreter or not, though. I’m not sure how complex to go here. I do think that

pyenv global

not causing Pants to change behavior is pretty confusing for a user

the versions selected by

pyenv global

are stored in

$PYENV_ROOT/version