##// END OF EJS Templates
procutil: avoid using os.fork() to implement runbgcommand...
procutil: avoid using os.fork() to implement runbgcommand We ran into the following deadlock: - some command creates an ssh peer, then raises without explicitly closing the peer (hg id + extension in our case) - dispatch catches the exception, calls ui.log('commandfinish', ..) (the sshpeer is still not closed), which calls logtoprocess, which calls procutil.runbgcommand. - in the child of runbgcommand's fork(), between the fork and the exec, the opening of file descriptors triggers a gc which runs the destructor for sshpeer, which waits on ssh's stderr being closed, which never happens since ssh's stderr is held open by the parent of the fork where said destructor hasn't run Remotefilelog appears to have a hack around this deadlock as well. I don't know if there's more subtlety to it, because even though the problem is determistic, it is very fragile, so I didn't manage to reduce it. I can imagine three ways of tackling this problem: 1. don't run any python between fork and exec in runbgcommand 2. make the finalizer harmless after the fork 3. close the peer without relying on gc behavior This commit goes with 1, as forking without exec'ing is tricky in general in a language with gc finalizers. And maybe it's better in the presence of rust threads. A future commit will try 2 or 3. Performance wise: at low memory usage, it's an improvement. At higher memory usage, it's about 2x faster than before when ensurestart=True, but 2x slower when ensurestart=False. Not sure if that matters. The reason for that last bit is that the subprocess.Popen always waits for the execve to finish, and at high memory usage, execve is slow because it deallocates the large page table. Numbers and script: before after mem=1.0GB, ensurestart=True 52.1ms 26.0ms mem=1.0GB, ensurestart=False 14.7ms 26.0ms mem=0.5GB, ensurestart=True 23.2ms 11.2ms mem=0.5GB, ensurestart=False 6.2ms 11.3ms mem=0.2GB, ensurestart=True 15.7ms 7.4ms mem=0.2GB, ensurestart=False 4.3ms 8.1ms mem=0.0GB, ensurestart=True 2.3ms 0.7ms mem=0.0GB, ensurestart=False 0.8ms 0.8ms import time for memsize in [1_000_000_000, 500_000_000, 250_000_000, 0]: mem = 'a' * memsize for ensurestart in [True, False]: now = time.time() n = 100 for i in range(n): procutil.runbgcommand([b'true'], {}, ensurestart=ensurestart) after = time.time() ms = (after - now) / float(n) * 1000 print(f'mem={memsize / 1e9:.1f}GB, ensurestart={ensurestart} -> {ms:.1f}ms') Differential Revision: https://phab.mercurial-scm.org/D9019

File last commit:

r47533:c94fa884 default
r47651:8759e22f default
Show More
files.rs
463 lines | 13.7 KiB | application/rls-services+xml | RustLexer
// files.rs
//
// Copyright 2019
// Raphaël Gomès <rgomes@octobus.net>,
// Yuya Nishihara <yuya@tcha.org>
//
// This software may be used and distributed according to the terms of the
// GNU General Public License version 2 or any later version.
//! Functions for fiddling with files.
use crate::utils::{
hg_path::{path_to_hg_path_buf, HgPath, HgPathBuf, HgPathError},
path_auditor::PathAuditor,
replace_slice,
};
use lazy_static::lazy_static;
use same_file::is_same_file;
use std::borrow::{Cow, ToOwned};
use std::ffi::OsStr;
use std::fs::Metadata;
use std::iter::FusedIterator;
use std::ops::Deref;
use std::path::{Path, PathBuf};
pub fn get_os_str_from_bytes(bytes: &[u8]) -> &OsStr {
let os_str;
#[cfg(unix)]
{
use std::os::unix::ffi::OsStrExt;
os_str = std::ffi::OsStr::from_bytes(bytes);
}
// TODO Handle other platforms
// TODO: convert from WTF8 to Windows MBCS (ANSI encoding).
// Perhaps, the return type would have to be Result<PathBuf>.
os_str
}
pub fn get_path_from_bytes(bytes: &[u8]) -> &Path {
Path::new(get_os_str_from_bytes(bytes))
}
// TODO: need to convert from WTF8 to MBCS bytes on Windows.
// that's why Vec<u8> is returned.
#[cfg(unix)]
pub fn get_bytes_from_path(path: impl AsRef<Path>) -> Vec<u8> {
get_bytes_from_os_str(path.as_ref())
}
#[cfg(unix)]
pub fn get_bytes_from_os_str(str: impl AsRef<OsStr>) -> Vec<u8> {
use std::os::unix::ffi::OsStrExt;
str.as_ref().as_bytes().to_vec()
}
/// An iterator over repository path yielding itself and its ancestors.
#[derive(Copy, Clone, Debug)]
pub struct Ancestors<'a> {
next: Option<&'a HgPath>,
}
impl<'a> Iterator for Ancestors<'a> {
type Item = &'a HgPath;
fn next(&mut self) -> Option<Self::Item> {
let next = self.next;
self.next = match self.next {
Some(s) if s.is_empty() => None,
Some(s) => {
let p = s.bytes().rposition(|c| *c == b'/').unwrap_or(0);
Some(HgPath::new(&s.as_bytes()[..p]))
}
None => None,
};
next
}
}
impl<'a> FusedIterator for Ancestors<'a> {}
/// An iterator over repository path yielding itself and its ancestors.
#[derive(Copy, Clone, Debug)]
pub(crate) struct AncestorsWithBase<'a> {
next: Option<(&'a HgPath, &'a HgPath)>,
}
impl<'a> Iterator for AncestorsWithBase<'a> {
type Item = (&'a HgPath, &'a HgPath);
fn next(&mut self) -> Option<Self::Item> {
let next = self.next;
self.next = match self.next {
Some((s, _)) if s.is_empty() => None,
Some((s, _)) => Some(s.split_filename()),
None => None,
};
next
}
}
impl<'a> FusedIterator for AncestorsWithBase<'a> {}
/// Returns an iterator yielding ancestor directories of the given repository
/// path.
///
/// The path is separated by '/', and must not start with '/'.
///
/// The path itself isn't included unless it is b"" (meaning the root
/// directory.)
pub fn find_dirs(path: &HgPath) -> Ancestors {
let mut dirs = Ancestors { next: Some(path) };
if !path.is_empty() {
dirs.next(); // skip itself
}
dirs
}
/// Returns an iterator yielding ancestor directories of the given repository
/// path.
///
/// The path is separated by '/', and must not start with '/'.
///
/// The path itself isn't included unless it is b"" (meaning the root
/// directory.)
pub(crate) fn find_dirs_with_base(path: &HgPath) -> AncestorsWithBase {
let mut dirs = AncestorsWithBase {
next: Some((path, HgPath::new(b""))),
};
if !path.is_empty() {
dirs.next(); // skip itself
}
dirs
}
/// TODO more than ASCII?
pub fn normalize_case(path: &HgPath) -> HgPathBuf {
#[cfg(windows)] // NTFS compares via upper()
return path.to_ascii_uppercase();
#[cfg(unix)]
path.to_ascii_lowercase()
}
lazy_static! {
static ref IGNORED_CHARS: Vec<Vec<u8>> = {
[
0x200c, 0x200d, 0x200e, 0x200f, 0x202a, 0x202b, 0x202c, 0x202d,
0x202e, 0x206a, 0x206b, 0x206c, 0x206d, 0x206e, 0x206f, 0xfeff,
]
.iter()
.map(|code| {
std::char::from_u32(*code)
.unwrap()
.encode_utf8(&mut [0; 3])
.bytes()
.collect()
})
.collect()
};
}
fn hfs_ignore_clean(bytes: &[u8]) -> Vec<u8> {
let mut buf = bytes.to_owned();
let needs_escaping = bytes.iter().any(|b| *b == b'\xe2' || *b == b'\xef');
if needs_escaping {
for forbidden in IGNORED_CHARS.iter() {
replace_slice(&mut buf, forbidden, &[])
}
buf
} else {
buf
}
}
pub fn lower_clean(bytes: &[u8]) -> Vec<u8> {
hfs_ignore_clean(&bytes.to_ascii_lowercase())
}
#[derive(Eq, PartialEq, Ord, PartialOrd, Copy, Clone)]
pub struct HgMetadata {
pub st_dev: u64,
pub st_mode: u32,
pub st_nlink: u64,
pub st_size: u64,
pub st_mtime: i64,
pub st_ctime: i64,
}
// TODO support other plaforms
#[cfg(unix)]
impl HgMetadata {
pub fn from_metadata(metadata: Metadata) -> Self {
use std::os::unix::fs::MetadataExt;
Self {
st_dev: metadata.dev(),
st_mode: metadata.mode(),
st_nlink: metadata.nlink(),
st_size: metadata.size(),
st_mtime: metadata.mtime(),
st_ctime: metadata.ctime(),
}
}
pub fn is_symlink(&self) -> bool {
// This is way too manual, but `HgMetadata` will go away in the
// near-future dirstate rewrite anyway.
self.st_mode & 0170000 == 0120000
}
}
/// Returns the canonical path of `name`, given `cwd` and `root`
pub fn canonical_path(
root: impl AsRef<Path>,
cwd: impl AsRef<Path>,
name: impl AsRef<Path>,
) -> Result<PathBuf, HgPathError> {
// TODO add missing normalization for other platforms
let root = root.as_ref();
let cwd = cwd.as_ref();
let name = name.as_ref();
let name = if !name.is_absolute() {
root.join(&cwd).join(&name)
} else {
name.to_owned()
};
let auditor = PathAuditor::new(&root);
if name != root && name.starts_with(&root) {
let name = name.strip_prefix(&root).unwrap();
auditor.audit_path(path_to_hg_path_buf(name)?)?;
Ok(name.to_owned())
} else if name == root {
Ok("".into())
} else {
// Determine whether `name' is in the hierarchy at or beneath `root',
// by iterating name=name.parent() until it returns `None` (can't
// check name == '/', because that doesn't work on windows).
let mut name = name.deref();
let original_name = name.to_owned();
loop {
let same = is_same_file(&name, &root).unwrap_or(false);
if same {
if name == original_name {
// `name` was actually the same as root (maybe a symlink)
return Ok("".into());
}
// `name` is a symlink to root, so `original_name` is under
// root
let rel_path = original_name.strip_prefix(&name).unwrap();
auditor.audit_path(path_to_hg_path_buf(&rel_path)?)?;
return Ok(rel_path.to_owned());
}
name = match name.parent() {
None => break,
Some(p) => p,
};
}
// TODO hint to the user about using --cwd
// Bubble up the responsibility to Python for now
Err(HgPathError::NotUnderRoot {
path: original_name.to_owned(),
root: root.to_owned(),
})
}
}
/// Returns the representation of the path relative to the current working
/// directory for display purposes.
///
/// `cwd` is a `HgPath`, so it is considered relative to the root directory
/// of the repository.
///
/// # Examples
///
/// ```
/// use hg::utils::hg_path::HgPath;
/// use hg::utils::files::relativize_path;
/// use std::borrow::Cow;
///
/// let file = HgPath::new(b"nested/file");
/// let cwd = HgPath::new(b"");
/// assert_eq!(relativize_path(file, cwd), Cow::Borrowed(b"nested/file"));
///
/// let cwd = HgPath::new(b"nested");
/// assert_eq!(relativize_path(file, cwd), Cow::Borrowed(b"file"));
///
/// let cwd = HgPath::new(b"other");
/// assert_eq!(relativize_path(file, cwd), Cow::Borrowed(b"../nested/file"));
/// ```
pub fn relativize_path(path: &HgPath, cwd: impl AsRef<HgPath>) -> Cow<[u8]> {
if cwd.as_ref().is_empty() {
Cow::Borrowed(path.as_bytes())
} else {
// This is not all accurate as to how large `res` will actually be, but
// profiling `rhg files` on a large-ish repo shows it’s better than
// starting from a zero-capacity `Vec` and letting `extend` reallocate
// repeatedly.
let guesstimate = path.as_bytes().len();
let mut res: Vec<u8> = Vec::with_capacity(guesstimate);
let mut path_iter = path.as_bytes().split(|b| *b == b'/').peekable();
let mut cwd_iter =
cwd.as_ref().as_bytes().split(|b| *b == b'/').peekable();
loop {
match (path_iter.peek(), cwd_iter.peek()) {
(Some(a), Some(b)) if a == b => (),
_ => break,
}
path_iter.next();
cwd_iter.next();
}
let mut need_sep = false;
for _ in cwd_iter {
if need_sep {
res.extend(b"/")
} else {
need_sep = true
};
res.extend(b"..");
}
for c in path_iter {
if need_sep {
res.extend(b"/")
} else {
need_sep = true
};
res.extend(c);
}
Cow::Owned(res)
}
}
#[cfg(test)]
mod tests {
use super::*;
use pretty_assertions::assert_eq;
#[test]
fn find_dirs_some() {
let mut dirs = super::find_dirs(HgPath::new(b"foo/bar/baz"));
assert_eq!(dirs.next(), Some(HgPath::new(b"foo/bar")));
assert_eq!(dirs.next(), Some(HgPath::new(b"foo")));
assert_eq!(dirs.next(), Some(HgPath::new(b"")));
assert_eq!(dirs.next(), None);
assert_eq!(dirs.next(), None);
}
#[test]
fn find_dirs_empty() {
// looks weird, but mercurial.pathutil.finddirs(b"") yields b""
let mut dirs = super::find_dirs(HgPath::new(b""));
assert_eq!(dirs.next(), Some(HgPath::new(b"")));
assert_eq!(dirs.next(), None);
assert_eq!(dirs.next(), None);
}
#[test]
fn test_find_dirs_with_base_some() {
let mut dirs = super::find_dirs_with_base(HgPath::new(b"foo/bar/baz"));
assert_eq!(
dirs.next(),
Some((HgPath::new(b"foo/bar"), HgPath::new(b"baz")))
);
assert_eq!(
dirs.next(),
Some((HgPath::new(b"foo"), HgPath::new(b"bar")))
);
assert_eq!(dirs.next(), Some((HgPath::new(b""), HgPath::new(b"foo"))));
assert_eq!(dirs.next(), None);
assert_eq!(dirs.next(), None);
}
#[test]
fn test_find_dirs_with_base_empty() {
let mut dirs = super::find_dirs_with_base(HgPath::new(b""));
assert_eq!(dirs.next(), Some((HgPath::new(b""), HgPath::new(b""))));
assert_eq!(dirs.next(), None);
assert_eq!(dirs.next(), None);
}
#[test]
fn test_canonical_path() {
let root = Path::new("/repo");
let cwd = Path::new("/dir");
let name = Path::new("filename");
assert_eq!(
canonical_path(root, cwd, name),
Err(HgPathError::NotUnderRoot {
path: PathBuf::from("/dir/filename"),
root: root.to_path_buf()
})
);
let root = Path::new("/repo");
let cwd = Path::new("/");
let name = Path::new("filename");
assert_eq!(
canonical_path(root, cwd, name),
Err(HgPathError::NotUnderRoot {
path: PathBuf::from("/filename"),
root: root.to_path_buf()
})
);
let root = Path::new("/repo");
let cwd = Path::new("/");
let name = Path::new("repo/filename");
assert_eq!(
canonical_path(root, cwd, name),
Ok(PathBuf::from("filename"))
);
let root = Path::new("/repo");
let cwd = Path::new("/repo");
let name = Path::new("filename");
assert_eq!(
canonical_path(root, cwd, name),
Ok(PathBuf::from("filename"))
);
let root = Path::new("/repo");
let cwd = Path::new("/repo/subdir");
let name = Path::new("filename");
assert_eq!(
canonical_path(root, cwd, name),
Ok(PathBuf::from("subdir/filename"))
);
}
#[test]
fn test_canonical_path_not_rooted() {
use std::fs::create_dir;
use tempfile::tempdir;
let base_dir = tempdir().unwrap();
let base_dir_path = base_dir.path();
let beneath_repo = base_dir_path.join("a");
let root = base_dir_path.join("a/b");
let out_of_repo = base_dir_path.join("c");
let under_repo_symlink = out_of_repo.join("d");
create_dir(&beneath_repo).unwrap();
create_dir(&root).unwrap();
// TODO make portable
std::os::unix::fs::symlink(&root, &out_of_repo).unwrap();
assert_eq!(
canonical_path(&root, Path::new(""), out_of_repo),
Ok(PathBuf::from(""))
);
assert_eq!(
canonical_path(&root, Path::new(""), &beneath_repo),
Err(HgPathError::NotUnderRoot {
path: beneath_repo.to_owned(),
root: root.to_owned()
})
);
assert_eq!(
canonical_path(&root, Path::new(""), &under_repo_symlink),
Ok(PathBuf::from("d"))
);
}
}