proposal: syscall: implement Flock on Solaris in terms of FcntlFlock

[ppbrown]

go version go1.10.1 solaris/amd64

okay I read that
"This package is locked down. Code outside the standard Go repository should be migrated to use the corresponding package in the golang.org/x/sys repository"
I get it. You dont want new stuff there.

That being said, programs are still going to be using syscall.Flock for years to come
Solaris doesnt have a Flock implementation, because current common implementation just uses Syscall(), which solaris doesnt exactly have.

But it DOES have FcntlFlock.

So would it be permissible for me to submit a patch to syscall_solaris.go, that gives Solaris an implementation of Flock on top of FcntlFlock ?

[bcmills]

The syscall package is for operating system primitives. If flock is not a primitive on Solaris, that doesn't seem like a good fit. (In particular, it appears that fcntl differs from flock in some subtle but significant ways.)

The os package contains non‐primitive file operations. Is there a higher‐level file locking API that could fit there? (Is there an analogous API on Windows?)

[rsc]

implement Flock on Solaris in terms of FcntlFlock

These are two semantically different locking APIs. I don't see why we'd implement one in terms of the other. Syscall is about raw access to the underlying system, not simulations.

In any event, x/sys/unix is the place to add both, and it looks like it already has both.

syscall is frozen, as others have said. We'd only add to syscall if uses within the standard library were affected, and I don't see any.