-
Notifications
You must be signed in to change notification settings - Fork 13.9k
motor: Use UTF-8 guarantee for OS strings #147797
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
|
rustbot has assigned @Mark-Simulacrum. Use |
|
My thinking during review was indeed that these should be zero-cost, but I missed that it used the |
|
I like that approach better and will implement it soon |
|
Superseded by #147932 |
Create UTF-8 version of `OsStr`/`OsString` Implement a UTF-8 version of `OsStr`/`OsString`, in addition to the existing bytes and WTF-8 platform-dependent encodings. This is applicable for several platforms, but I've currently only implemented it for Motor OS: - WASI uses Unicode paths, but currently reexports the Unix bytes-assuming `OsStrExt`/`OsStringExt` traits. - [wasi:filesystem](https://wa.dev/wasi:filesystem) APIs: > Paths are passed as interface-type `strings`, meaning they must consist of a sequence of Unicode Scalar Values (USVs). Some filesystems may contain paths which are not accessible by this API. - In [wasi-filesystem#17](WebAssembly/wasi-filesystem#17 (comment)), it was decided that applications can use any Unicode transformation format, so we're free to use UTF-8 (and probably already do). This was chosen over specifically UTF-8 or an ad hoc encoding which preserves paths not representable in UTF-8. > The current API uses strings for filesystem paths, which contains sequences of Unicode scalar values (USVs), which applications can work with using strings encoded in UTF-8, UTF-16, or other Unicode encodings. > > This does mean that the API is unable to open files which do not have well-formed Unicode encodings, which may want separate APIs for handling such paths or may want something like the arf-strings proposal, but if we need that we should file a new issue for it. - As of Redox OS [0.7.0](https://www.redox-os.org/news/release-0.7.0/), "All paths are now required to be UTF-8, and the kernel enforces this". This appears to have been implemented in commit [d331f72f](https://gitlab.redox-os.org/redox-os/kernel/-/commit/d331f72f2a51fa577072f24bc2587829fd87368b) (Use UTF-8 for all paths, 2021-02-14). Redox does not have `OsStrExt`/`OsStringExt`. - Motor OS guarantees that its OS strings are UTF-8 in its [current `OsStrExt`/`OsStringExt` traits](https://github.com/moturus/rust/blob/a828ffcf5f04be5cdd91b5fad608102eabc17ec7/library/std/src/os/motor/ffi.rs), but they're still internally bytes like Unix. This is an alternate approach to rust-lang#147797, which reuses the existing bytes `OsString` and relies on the safety properties of `from_encoded_bytes_unchecked`. Compared to that, this also gains efficiency from propagating the UTF-8 invariant to the whole type, as it never needs to test for UTF-8 validity. Note that Motor OS currently does not build until rust-lang#147930 merges. cc `@tgross35` (for earlier review) cc `@alexcrichton,` `@rylev,` `@loganek` (for WASI) cc `@lasiotus` (for Motor OS) cc `@jackpot51` (for Redox OS)
Create UTF-8 version of `OsStr`/`OsString` Implement a UTF-8 version of `OsStr`/`OsString`, in addition to the existing bytes and WTF-8 platform-dependent encodings. This is applicable for several platforms, but I've currently only implemented it for Motor OS: - WASI uses Unicode paths, but currently reexports the Unix bytes-assuming `OsStrExt`/`OsStringExt` traits. - [wasi:filesystem](https://wa.dev/wasi:filesystem) APIs: > Paths are passed as interface-type `strings`, meaning they must consist of a sequence of Unicode Scalar Values (USVs). Some filesystems may contain paths which are not accessible by this API. - In [wasi-filesystem#17](WebAssembly/wasi-filesystem#17 (comment)), it was decided that applications can use any Unicode transformation format, so we're free to use UTF-8 (and probably already do). This was chosen over specifically UTF-8 or an ad hoc encoding which preserves paths not representable in UTF-8. > The current API uses strings for filesystem paths, which contains sequences of Unicode scalar values (USVs), which applications can work with using strings encoded in UTF-8, UTF-16, or other Unicode encodings. > > This does mean that the API is unable to open files which do not have well-formed Unicode encodings, which may want separate APIs for handling such paths or may want something like the arf-strings proposal, but if we need that we should file a new issue for it. - As of Redox OS [0.7.0](https://www.redox-os.org/news/release-0.7.0/), "All paths are now required to be UTF-8, and the kernel enforces this". This appears to have been implemented in commit [d331f72f](https://gitlab.redox-os.org/redox-os/kernel/-/commit/d331f72f2a51fa577072f24bc2587829fd87368b) (Use UTF-8 for all paths, 2021-02-14). Redox does not have `OsStrExt`/`OsStringExt`. - Motor OS guarantees that its OS strings are UTF-8 in its [current `OsStrExt`/`OsStringExt` traits](https://github.com/moturus/rust/blob/a828ffcf5f04be5cdd91b5fad608102eabc17ec7/library/std/src/os/motor/ffi.rs), but they're still internally bytes like Unix. This is an alternate approach to rust-lang/rust#147797, which reuses the existing bytes `OsString` and relies on the safety properties of `from_encoded_bytes_unchecked`. Compared to that, this also gains efficiency from propagating the UTF-8 invariant to the whole type, as it never needs to test for UTF-8 validity. Note that Motor OS currently does not build until rust-lang/rust#147930 merges. cc `@tgross35` (for earlier review) cc `@alexcrichton,` `@rylev,` `@loganek` (for WASI) cc `@lasiotus` (for Motor OS) cc `@jackpot51` (for Redox OS)
Motor OS (a new target added in #147000) guarantees that strings from the OS are valid UTF-8, yet the
OsStrExt/OsStringExttraits, which aim to expose the inner UTF-8 representations, use checked UTF-8 conversions.The only way for a user to construct an
OsStr/OsStringfrom arbitrary bytes is viafrom_encoded_bytes_unchecked, which requires:Thus, callers are required to supply bytes which originated from
OsStr::as_encoded_bytes(i.e., guaranteed to be UTF-8) and/or are valid UTF-8, so it is library UB for anOsStr/OsStringto contain invalid UTF-8 on Motor OS.Since the standard library can make these guarantees, I think it is appropriate for these extension traits to perform unchecked conversions. As they stand, they offer no benefit over existing methods.
Also: Replace
OsStringExt::as_strwithOsStringExt::into_string(just useOsStr::as_strfor the other) and mirror the Unix comments for these functions.cc @lasiotus