'chmod' equivalent in snapshot preview1?

[titzer]

It appears that snapshot preview 1 does not have a way to change file permissions. This is important functionality for one of my use cases (a compiler that generates executables). This functionality does appear to be planned for the next version of wasi.

Is snapshot preview1 being updated? Is there interest in adding a chmod-like capability?

If so, I would propose a simple fs_chmod(fd: i32, mode: i32) function.

[sunfishcode]

We've have a permissions_set function in the latest proposal, though it's not yet in a published shapshot. It's currently limited to just read/write/execute, as well as a private flag, rather than exposing the Unix user/group/other bits individually. But the upside is that it should be implementable on Windows. Do you think that would be enough for your usecase, or do you need more fine-grained control?

Adding a full chmod is also an option, though it might need to be designed in a way that would allow it to fail if the underlying filesystem doesn't use Unix-style user/group/other.

[titzer]

For my use case, just u+x would be enough to limp along, but that feels like it wouldn't be general enough. Designating permission categories user/group/other and failing if one of the categories does not exist would allow an app to at least try setting permissions. Keep in mind there is also the setuid bit, which is esoteric, but useful in some situations.

[sunfishcode]

setuid is tricky (even ignoring its long troubled security history), because setuid allows programs to effectively grant the full host-process authority to other programs, and WASI programs may be considered less privileged than the host-process user. There are ways we could model it, such as passing programs an extra handle to represent a "host-process authority" that they'd have to pass into the API to set a setuid bit. But before we start designing something, it'd be good to have specific motivating use cases.

[titzer]

I'm happy to leave setuid out of functionality for chmod, as I don't have a use case for it. I just mentioned it for completeness.