impl Default for ExitStatus

[ijackson]

Proposal

Problem statement

It is not possible for code outside std to portably create an ExitStatus without actually running a process.

Motivation, use-cases

• You might want to create a value as a placeholder in error handling APIs (for example, Command improvements for ergonomics and error handling rfcs#3362 https://github.com/ijackson/rust-rfcs/blob/command/text/0000-command-ergonomics.md#more-convenient-way-to-run-diff)
• Testing

In some cases it might be convenient to use Option<ExitStatus>; but ExitStatus is already kind of option-ish in that it can represent an error, or no error.

It is IMO not reasonable to have a type that a user can mostly-portably create, but to require a user to run a dummy process to do so. This type can already be non-portably constructed on Unix and Windows via ExitStatusExt.

Solution sketches

The obvious-to-me answer is to impl Default for ExitStatus.

I did that in rust-lang/rust#106425 and am now belatedly making this ACP.

A consequence is that ExitStatus now becomes inhabited even on platforms without subprocesses. This seems reasonable.

Links and related work

See also the tracking issue for thec currently-unstable ExitStatusError, rust-lang/rust#84908. I'm postponing addressing the same question for ExitStatusError, but it should be considered before stabilisation.

What happens now?

This issue is part of the libs-api team API change proposal process. Once this issue is filed the libs-api team will review open proposals in its weekly meeting. You should receive feedback within a week or two.

[m-ou-se]

Thanks! This seems fine to me. I'll start the FCP on rust-lang/rust#106425.

(Perhaps we should also allow the creation of other statusses, maybe in the form of a From<ExitCode>. Although maybe ExitStatusExt::from_raw already suffices, at least on Windows and Unix platforms.)