##// END OF EJS Templates
hgweb: encode WSGI environment using the ISO-8859-1 codec...
hgweb: encode WSGI environment using the ISO-8859-1 codec The WSGI specification (PEP 3333) specifies that on Python 3 all strings passed by the server must be of type str with code points encodable using the ISO 8859-1 codec. For some reason, I introduced a bug in 2632c1ed8f34 by applying the reverse change. Maybe I got confused because PEP 3333 says that arbitrary operating system environment variables may be contained in the WSGI environment and therefore we need to handle the WSGI environment variables like we would handle operating system environment variables. The bug mentioned in the previous paragraph and fixed by this changeset manifested e.g. in the path of the URL being encoded in the wrong way. Browsers encode non-ASCII bytes with the percent-encoding. WSGI servers will decode the percent-encoded bytes and pass them to the application as strings where each byte is mapped to the corresponding code point with the same ordinal (i.e. it is decoded using the ISO-8859-1 codec). Mercurial uses the bytes type for these strings (which makes much more sense), so we need to encode it again using the ISO-8859-1 codec. If we use another codec, it can result in nonsense.

File last commit:

r50828:be3b545c default
r51713:9ed281bb stable
Show More
pybytes_deref.rs
57 lines | 1.6 KiB | application/rls-services+xml | RustLexer
use cpython::{PyBytes, Python};
use stable_deref_trait::StableDeref;
/// Safe abstraction over a `PyBytes` together with the `&[u8]` slice
/// that borrows it. Implements `Deref<Target = [u8]>`.
///
/// Calling `PyBytes::data` requires a GIL marker but we want to access the
/// data in a thread that (ideally) does not need to acquire the GIL.
/// This type allows separating the call an the use.
///
/// It also enables using a (wrapped) `PyBytes` in GIL-unaware generic code.
pub struct PyBytesDeref {
#[allow(unused)]
keep_alive: PyBytes,
/// Borrows the buffer inside `self.keep_alive`,
/// but the borrow-checker cannot express self-referential structs.
data: *const [u8],
}
impl PyBytesDeref {
pub fn new(py: Python, bytes: PyBytes) -> Self {
Self {
data: bytes.data(py),
keep_alive: bytes,
}
}
pub fn unwrap(self) -> PyBytes {
self.keep_alive
}
}
impl std::ops::Deref for PyBytesDeref {
type Target = [u8];
fn deref(&self) -> &[u8] {
// Safety: the raw pointer is valid as long as the PyBytes is still
// alive, and the returned slice borrows `self`.
unsafe { &*self.data }
}
}
unsafe impl StableDeref for PyBytesDeref {}
fn require_send<T: Send>() {}
#[allow(unused)]
fn static_assert_pybytes_is_send() {
#[allow(clippy::no_effect)]
require_send::<PyBytes>;
}
// Safety: PyBytes is Send. Raw pointers are not by default,
// but here sending one to another thread is fine since we ensure it stays
// valid.
unsafe impl Send for PyBytesDeref {}