Tracking Issue for NUL-terminated file names with `#[track_caller]`

[Darksonn]

Feature gate: #![feature(file_with_nul)]

This is a tracking issue for Location::file_with_nul.

This feature allows you to obtain NUL-terminated file names from core::panic::Location when using #[track_caller]. This allows for better error messages in projects performing interop with C/C++.

Public API

// core::panic

pub struct Location;

impl Location {
    pub const fn file_with_nul(&self) -> &CStr;
}

Steps / History

(Remember to update the S-tracking-* label when checking boxes.)

• Accept ACP: ACP: Add nul-terminated version of core::panic::Location::file libs-team#466
• Implementation:
  • Add Location::file_with_nul #135054
• Resolve open questions:
  • Use raw pointer for NUL-terminated file location with #[track_caller] #142579
• Final comment period (FCP)¹
• Stabilization PR
  • Stabilize std::panic::Location::file_with_nul #145664

Related PRs:

• Add nul-terminated filename for #[track_caller] #131828
• Add unstable option to nul-terminate location strings #135240
• Null terminate core::panic::Location file strings #117431
• Use raw pointer for NUL-terminated file location with #[track_caller] #142579
• Do not include NUL-terminator in computed length #142708

Footnotes

1. https://std-dev-guide.rust-lang.org/feature-lifecycle/stabilization.html ↩

[bonzini]

As a possible user of the feature, returning *const c_char would be just fine. I have no use for the stored file other than at the Rust->C boundary, and would use file.as_ptr() there.

The only minor nuisance of storing a *const instead of a CStr is that it blocks automatic implementation of Send and Sync.

[bjorn3]

We need to store the length anyway for Location::file(), so returning &CStr from Location::file_with_nul shouldn't increase the size of Location at all.

[Darksonn]

@bjorn3 That's not true. Check out #117431 which used strlen every time Location::file() is called to compute the length in exchange for reducing the size of Location.