view rust/hg-core/src/filepatterns.rs @ 45077:fa270dcbdb55

procutil: back out 8403cc54bc83 (make ....procutil.stderr unbuffered)

Changeset 8403cc54bc83 introduced code that opens a second file object
referring to the stderr file descriptor. This broke tests on Windows. The
reason is that on Windows, sys.stderr is buffered and procutil.stderr closed
the file descriptor when it got garbage collected before sys.stderr had the
chance to flush buffered data.

`procutil.stdout` had the same problem for a long time, but we didn’t realize,
as in CI test runs, stdout is not a TTY and in this case no second file object
is opened.

// filepatterns.rs
//
// Copyright 2019 Raphaël Gomès <[e-mail]>
//
// This software may be used and distributed according to the terms of the
// GNU General Public License version 2 or any later version.

//! Handling of Mercurial-specific patterns.

use crate::{
    utils::{
        files::{canonical_path, get_bytes_from_path, get_path_from_bytes},
        hg_path::{path_to_hg_path_buf, HgPathBuf, HgPathError},
        SliceExt,
    },
    FastHashMap, PatternError,
};
use lazy_static::lazy_static;
use regex::bytes::{NoExpand, Regex};
use std::fs::File;
use std::io::Read;
use std::ops::Deref;
use std::path::{Path, PathBuf};
use std::vec::Vec;

lazy_static! {
    static ref RE_ESCAPE: Vec<Vec<u8>> = {
        let mut v: Vec<Vec<u8>> = (0..=255).map(|byte| vec![byte]).collect();
        let to_escape = b"()[]{}?*+-|^$\\.&~# \t\n\r\x0b\x0c";
        for byte in to_escape {
            v[*byte as usize].insert(0, b'\\');
        }
        v
    };
}

/// These are matched in order
const GLOB_REPLACEMENTS: &[(&[u8], &[u8])] =
    &[(b"*/", b"(?:.*/)?"), (b"*", b".*"), (b"", b"[^/]*")];

/// Appended to the regexp of globs
const GLOB_SUFFIX: &[u8; 7] = b"(?:/|$)";

#[derive(Debug, Copy, Clone, PartialEq, Eq)]
pub enum PatternSyntax {
    /// A regular expression
    Regexp,
    /// Glob that matches at the front of the path
    RootGlob,
    /// Glob that matches at any suffix of the path (still anchored at
    /// slashes)
    Glob,
    /// a path relative to repository root, which is matched recursively
    Path,
    /// A path relative to cwd
    RelPath,
    /// an unrooted glob (*.rs matches Rust files in all dirs)
    RelGlob,
    /// A regexp that needn't match the start of a name
    RelRegexp,
    /// A path relative to repository root, which is matched non-recursively
    /// (will not match subdirectories)
    RootFiles,
    /// A file of patterns to read and include
    Include,
    /// A file of patterns to match against files under the same directory
    SubInclude,
}

/// Transforms a glob pattern into a regex
fn glob_to_re(pat: &[u8]) -> Vec<u8> {
    let mut input = pat;
    let mut res: Vec<u8> = vec![];
    let mut group_depth = 0;

    while let Some((c, rest)) = input.split_first() {
        input = rest;

        match c {
            b'*' => {
                for (source, repl) in GLOB_REPLACEMENTS {
                    if let Some(rest) = input.drop_prefix(source) {
                        input = rest;
                        res.extend(*repl);
                        break;
                    }
                }
            }
            b'?' => res.extend(b"."),
            b'[' => {
                match input.iter().skip(1).position(|b| *b == b']') {
                    None => res.extend(b"\\["),
                    Some(end) => {
                        // Account for the one we skipped
                        let end = end + 1;

                        res.extend(b"[");

                        for (i, b) in input[..end].iter().enumerate() {
                            if *b == b'!' && i == 0 {
                                res.extend(b"^")
                            } else if *b == b'^' && i == 0 {
                                res.extend(b"\\^")
                            } else if *b == b'\\' {
                                res.extend(b"\\\\")
                            } else {
                                res.push(*b)
                            }
                        }
                        res.extend(b"]");
                        input = &input[end + 1..];
                    }
                }
            }
            b'{' => {
                group_depth += 1;
                res.extend(b"(?:")
            }
            b'}' if group_depth > 0 => {
                group_depth -= 1;
                res.extend(b")");
            }
            b',' if group_depth > 0 => res.extend(b"|"),
            b'\\' => {
                let c = {
                    if let Some((c, rest)) = input.split_first() {
                        input = rest;
                        c
                    } else {
                        c
                    }
                };
                res.extend(&RE_ESCAPE[*c as usize])
            }
            _ => res.extend(&RE_ESCAPE[*c as usize]),
        }
    }
    res
}

fn escape_pattern(pattern: &[u8]) -> Vec<u8> {
    pattern
        .iter()
        .flat_map(|c| RE_ESCAPE[*c as usize].clone())
        .collect()
}

pub fn parse_pattern_syntax(
    kind: &[u8],
) -> Result<PatternSyntax, PatternError> {
    match kind {
        b"re:" => Ok(PatternSyntax::Regexp),
        b"path:" => Ok(PatternSyntax::Path),
        b"relpath:" => Ok(PatternSyntax::RelPath),
        b"rootfilesin:" => Ok(PatternSyntax::RootFiles),
        b"relglob:" => Ok(PatternSyntax::RelGlob),
        b"relre:" => Ok(PatternSyntax::RelRegexp),
        b"glob:" => Ok(PatternSyntax::Glob),
        b"rootglob:" => Ok(PatternSyntax::RootGlob),
        b"include:" => Ok(PatternSyntax::Include),
        b"subinclude:" => Ok(PatternSyntax::SubInclude),
        _ => Err(PatternError::UnsupportedSyntax(
            String::from_utf8_lossy(kind).to_string(),
        )),
    }
}

/// Builds the regex that corresponds to the given pattern.
/// If within a `syntax: regexp` context, returns the pattern,
/// otherwise, returns the corresponding regex.
fn _build_single_regex(entry: &IgnorePattern) -> Vec<u8> {
    let IgnorePattern {
        syntax, pattern, ..
    } = entry;
    if pattern.is_empty() {
        return vec![];
    }
    match syntax {
        PatternSyntax::Regexp => pattern.to_owned(),
        PatternSyntax::RelRegexp => {
            // The `regex` crate accepts `**` while `re2` and Python's `re`
            // do not. Checking for `*` correctly triggers the same error all
            // engines.
            if pattern[0] == b'^'
                || pattern[0] == b'*'
                || pattern.starts_with(b".*")
            {
                return pattern.to_owned();
            }
            [&b".*"[..], pattern].concat()
        }
        PatternSyntax::Path | PatternSyntax::RelPath => {
            if pattern == b"." {
                return vec![];
            }
            [escape_pattern(pattern).as_slice(), b"(?:/|$)"].concat()
        }
        PatternSyntax::RootFiles => {
            let mut res = if pattern == b"." {
                vec![]
            } else {
                // Pattern is a directory name.
                [escape_pattern(pattern).as_slice(), b"/"].concat()
            };

            // Anything after the pattern must be a non-directory.
            res.extend(b"[^/]+$");
            res
        }
        PatternSyntax::RelGlob => {
            let glob_re = glob_to_re(pattern);
            if let Some(rest) = glob_re.drop_prefix(b"[^/]*") {
                [b".*", rest, GLOB_SUFFIX].concat()
            } else {
                [b"(?:.*/)?", glob_re.as_slice(), GLOB_SUFFIX].concat()
            }
        }
        PatternSyntax::Glob | PatternSyntax::RootGlob => {
            [glob_to_re(pattern).as_slice(), GLOB_SUFFIX].concat()
        }
        PatternSyntax::Include | PatternSyntax::SubInclude => unreachable!(),
    }
}

const GLOB_SPECIAL_CHARACTERS: [u8; 7] =
    [b'*', b'?', b'[', b']', b'{', b'}', b'\\'];

/// TODO support other platforms
#[cfg(unix)]
pub fn normalize_path_bytes(bytes: &[u8]) -> Vec<u8> {
    if bytes.is_empty() {
        return b".".to_vec();
    }
    let sep = b'/';

    let mut initial_slashes = bytes.iter().take_while(|b| **b == sep).count();
    if initial_slashes > 2 {
        // POSIX allows one or two initial slashes, but treats three or more
        // as single slash.
        initial_slashes = 1;
    }
    let components = bytes
        .split(|b| *b == sep)
        .filter(|c| !(c.is_empty() || c == b"."))
        .fold(vec![], |mut acc, component| {
            if component != b".."
                || (initial_slashes == 0 && acc.is_empty())
                || (!acc.is_empty() && acc[acc.len() - 1] == b"..")
            {
                acc.push(component)
            } else if !acc.is_empty() {
                acc.pop();
            }
            acc
        });
    let mut new_bytes = components.join(&sep);

    if initial_slashes > 0 {
        let mut buf: Vec<_> = (0..initial_slashes).map(|_| sep).collect();
        buf.extend(new_bytes);
        new_bytes = buf;
    }
    if new_bytes.is_empty() {
        b".".to_vec()
    } else {
        new_bytes
    }
}

/// Wrapper function to `_build_single_regex` that short-circuits 'exact' globs
/// that don't need to be transformed into a regex.
pub fn build_single_regex(
    entry: &IgnorePattern,
) -> Result<Option<Vec<u8>>, PatternError> {
    let IgnorePattern {
        pattern, syntax, ..
    } = entry;
    let pattern = match syntax {
        PatternSyntax::RootGlob
        | PatternSyntax::Path
        | PatternSyntax::RelGlob
        | PatternSyntax::RootFiles => normalize_path_bytes(&pattern),
        PatternSyntax::Include | PatternSyntax::SubInclude => {
            return Err(PatternError::NonRegexPattern(entry.clone()))
        }
        _ => pattern.to_owned(),
    };
    if *syntax == PatternSyntax::RootGlob
        && !pattern.iter().any(|b| GLOB_SPECIAL_CHARACTERS.contains(b))
    {
        Ok(None)
    } else {
        let mut entry = entry.clone();
        entry.pattern = pattern;
        Ok(Some(_build_single_regex(&entry)))
    }
}

lazy_static! {
    static ref SYNTAXES: FastHashMap<&'static [u8], &'static [u8]> = {
        let mut m = FastHashMap::default();

        m.insert(b"re".as_ref(), b"relre:".as_ref());
        m.insert(b"regexp".as_ref(), b"relre:".as_ref());
        m.insert(b"glob".as_ref(), b"relglob:".as_ref());
        m.insert(b"rootglob".as_ref(), b"rootglob:".as_ref());
        m.insert(b"include".as_ref(), b"include:".as_ref());
        m.insert(b"subinclude".as_ref(), b"subinclude:".as_ref());
        m
    };
}

#[derive(Debug)]
pub enum PatternFileWarning {
    /// (file path, syntax bytes)
    InvalidSyntax(PathBuf, Vec<u8>),
    /// File path
    NoSuchFile(PathBuf),
}

pub fn parse_pattern_file_contents<P: AsRef<Path>>(
    lines: &[u8],
    file_path: P,
    warn: bool,
) -> Result<(Vec<IgnorePattern>, Vec<PatternFileWarning>), PatternError> {
    let comment_regex = Regex::new(r"((?:^|[^\\])(?:\\\\)*)#.*").unwrap();

    #[allow(clippy::trivial_regex)]
    let comment_escape_regex = Regex::new(r"\\#").unwrap();
    let mut inputs: Vec<IgnorePattern> = vec![];
    let mut warnings: Vec<PatternFileWarning> = vec![];

    let mut current_syntax = b"relre:".as_ref();

    for (line_number, mut line) in lines.split(|c| *c == b'\n').enumerate() {
        let line_number = line_number + 1;

        let line_buf;
        if line.contains(&b'#') {
            if let Some(cap) = comment_regex.captures(line) {
                line = &line[..cap.get(1).unwrap().end()]
            }
            line_buf = comment_escape_regex.replace_all(line, NoExpand(b"#"));
            line = &line_buf;
        }

        let mut line = line.trim_end();

        if line.is_empty() {
            continue;
        }

        if let Some(syntax) = line.drop_prefix(b"syntax:") {
            let syntax = syntax.trim();

            if let Some(rel_syntax) = SYNTAXES.get(syntax) {
                current_syntax = rel_syntax;
            } else if warn {
                warnings.push(PatternFileWarning::InvalidSyntax(
                    file_path.as_ref().to_owned(),
                    syntax.to_owned(),
                ));
            }
            continue;
        }

        let mut line_syntax: &[u8] = &current_syntax;

        for (s, rels) in SYNTAXES.iter() {
            if let Some(rest) = line.drop_prefix(rels) {
                line_syntax = rels;
                line = rest;
                break;
            }
            if let Some(rest) = line.drop_prefix(&[s, &b":"[..]].concat()) {
                line_syntax = rels;
                line = rest;
                break;
            }
        }

        inputs.push(IgnorePattern::new(
            parse_pattern_syntax(&line_syntax).map_err(|e| match e {
                PatternError::UnsupportedSyntax(syntax) => {
                    PatternError::UnsupportedSyntaxInFile(
                        syntax,
                        file_path.as_ref().to_string_lossy().into(),
                        line_number,
                    )
                }
                _ => e,
            })?,
            &line,
            &file_path,
        ));
    }
    Ok((inputs, warnings))
}

pub fn read_pattern_file<P: AsRef<Path>>(
    file_path: P,
    warn: bool,
) -> Result<(Vec<IgnorePattern>, Vec<PatternFileWarning>), PatternError> {
    let mut f = match File::open(file_path.as_ref()) {
        Ok(f) => Ok(f),
        Err(e) => match e.kind() {
            std::io::ErrorKind::NotFound => {
                return Ok((
                    vec![],
                    vec![PatternFileWarning::NoSuchFile(
                        file_path.as_ref().to_owned(),
                    )],
                ))
            }
            _ => Err(e),
        },
    }?;
    let mut contents = Vec::new();

    f.read_to_end(&mut contents)?;

    Ok(parse_pattern_file_contents(&contents, file_path, warn)?)
}

/// Represents an entry in an "ignore" file.
#[derive(Debug, Eq, PartialEq, Clone)]
pub struct IgnorePattern {
    pub syntax: PatternSyntax,
    pub pattern: Vec<u8>,
    pub source: PathBuf,
}

impl IgnorePattern {
    pub fn new(
        syntax: PatternSyntax,
        pattern: &[u8],
        source: impl AsRef<Path>,
    ) -> Self {
        Self {
            syntax,
            pattern: pattern.to_owned(),
            source: source.as_ref().to_owned(),
        }
    }
}

pub type PatternResult<T> = Result<T, PatternError>;

/// Wrapper for `read_pattern_file` that also recursively expands `include:`
/// patterns.
///
/// `subinclude:` is not treated as a special pattern here: unraveling them
/// needs to occur in the "ignore" phase.
pub fn get_patterns_from_file(
    pattern_file: impl AsRef<Path>,
    root_dir: impl AsRef<Path>,
) -> PatternResult<(Vec<IgnorePattern>, Vec<PatternFileWarning>)> {
    let (patterns, mut warnings) = read_pattern_file(&pattern_file, true)?;
    let patterns = patterns
        .into_iter()
        .flat_map(|entry| -> PatternResult<_> {
            let IgnorePattern {
                syntax, pattern, ..
            } = &entry;
            Ok(match syntax {
                PatternSyntax::Include => {
                    let inner_include =
                        root_dir.as_ref().join(get_path_from_bytes(&pattern));
                    let (inner_pats, inner_warnings) = get_patterns_from_file(
                        &inner_include,
                        root_dir.as_ref(),
                    )?;
                    warnings.extend(inner_warnings);
                    inner_pats
                }
                _ => vec![entry],
            })
        })
        .flatten()
        .collect();

    Ok((patterns, warnings))
}

/// Holds all the information needed to handle a `subinclude:` pattern.
pub struct SubInclude {
    /// Will be used for repository (hg) paths that start with this prefix.
    /// It is relative to the current working directory, so comparing against
    /// repository paths is painless.
    pub prefix: HgPathBuf,
    /// The file itself, containing the patterns
    pub path: PathBuf,
    /// Folder in the filesystem where this it applies
    pub root: PathBuf,
}

impl SubInclude {
    pub fn new(
        root_dir: impl AsRef<Path>,
        pattern: &[u8],
        source: impl AsRef<Path>,
    ) -> Result<SubInclude, HgPathError> {
        let normalized_source =
            normalize_path_bytes(&get_bytes_from_path(source));

        let source_root = get_path_from_bytes(&normalized_source);
        let source_root =
            source_root.parent().unwrap_or_else(|| source_root.deref());

        let path = source_root.join(get_path_from_bytes(pattern));
        let new_root = path.parent().unwrap_or_else(|| path.deref());

        let prefix = canonical_path(&root_dir, &root_dir, new_root)?;

        Ok(Self {
            prefix: path_to_hg_path_buf(prefix).and_then(|mut p| {
                if !p.is_empty() {
                    p.push(b'/');
                }
                Ok(p)
            })?,
            path: path.to_owned(),
            root: new_root.to_owned(),
        })
    }
}

/// Separate and pre-process subincludes from other patterns for the "ignore"
/// phase.
pub fn filter_subincludes(
    ignore_patterns: &[IgnorePattern],
    root_dir: impl AsRef<Path>,
) -> Result<(Vec<SubInclude>, Vec<&IgnorePattern>), HgPathError> {
    let mut subincludes = vec![];
    let mut others = vec![];

    for ignore_pattern in ignore_patterns.iter() {
        let IgnorePattern {
            syntax,
            pattern,
            source,
        } = ignore_pattern;
        if *syntax == PatternSyntax::SubInclude {
            subincludes.push(SubInclude::new(&root_dir, pattern, &source)?);
        } else {
            others.push(ignore_pattern)
        }
    }
    Ok((subincludes, others))
}

#[cfg(test)]
mod tests {
    use super::*;
    use pretty_assertions::assert_eq;

    #[test]
    fn escape_pattern_test() {
        let untouched =
            br#"!"%',/0123456789:;<=>@ABCDEFGHIJKLMNOPQRSTUVWXYZ_`abcdefghijklmnopqrstuvwxyz"#;
        assert_eq!(escape_pattern(untouched), untouched.to_vec());
        // All escape codes
        assert_eq!(
            escape_pattern(br#"()[]{}?*+-|^$\\.&~# \t\n\r\v\f"#),
            br#"\(\)\[\]\{\}\?\*\+\-\|\^\$\\\\\.\&\~\#\ \\t\\n\\r\\v\\f"#
                .to_vec()
        );
    }

    #[test]
    fn glob_test() {
        assert_eq!(glob_to_re(br#"?"#), br#"."#);
        assert_eq!(glob_to_re(br#"*"#), br#"[^/]*"#);
        assert_eq!(glob_to_re(br#"**"#), br#".*"#);
        assert_eq!(glob_to_re(br#"**/a"#), br#"(?:.*/)?a"#);
        assert_eq!(glob_to_re(br#"a/**/b"#), br#"a/(?:.*/)?b"#);
        assert_eq!(glob_to_re(br#"[a*?!^][^b][!c]"#), br#"[a*?!^][\^b][^c]"#);
        assert_eq!(glob_to_re(br#"{a,b}"#), br#"(?:a|b)"#);
        assert_eq!(glob_to_re(br#".\*\?"#), br#"\.\*\?"#);
    }

    #[test]
    fn test_parse_pattern_file_contents() {
        let lines = b"syntax: glob\n*.elc";

        assert_eq!(
            parse_pattern_file_contents(lines, Path::new("file_path"), false)
                .unwrap()
                .0,
            vec![IgnorePattern::new(
                PatternSyntax::RelGlob,
                b"*.elc",
                Path::new("file_path")
            )],
        );

        let lines = b"syntax: include\nsyntax: glob";

        assert_eq!(
            parse_pattern_file_contents(lines, Path::new("file_path"), false)
                .unwrap()
                .0,
            vec![]
        );
        let lines = b"glob:**.o";
        assert_eq!(
            parse_pattern_file_contents(lines, Path::new("file_path"), false)
                .unwrap()
                .0,
            vec![IgnorePattern::new(
                PatternSyntax::RelGlob,
                b"**.o",
                Path::new("file_path")
            )]
        );
    }

    #[test]
    fn test_build_single_regex() {
        assert_eq!(
            build_single_regex(&IgnorePattern::new(
                PatternSyntax::RelGlob,
                b"rust/target/",
                Path::new("")
            ))
            .unwrap(),
            Some(br"(?:.*/)?rust/target(?:/|$)".to_vec()),
        );
        assert_eq!(
            build_single_regex(&IgnorePattern::new(
                PatternSyntax::Regexp,
                br"rust/target/\d+",
                Path::new("")
            ))
            .unwrap(),
            Some(br"rust/target/\d+".to_vec()),
        );
    }

    #[test]
    fn test_build_single_regex_shortcut() {
        assert_eq!(
            build_single_regex(&IgnorePattern::new(
                PatternSyntax::RootGlob,
                b"",
                Path::new("")
            ))
            .unwrap(),
            None,
        );
        assert_eq!(
            build_single_regex(&IgnorePattern::new(
                PatternSyntax::RootGlob,
                b"whatever",
                Path::new("")
            ))
            .unwrap(),
            None,
        );
        assert_eq!(
            build_single_regex(&IgnorePattern::new(
                PatternSyntax::RootGlob,
                b"*.o",
                Path::new("")
            ))
            .unwrap(),
            Some(br"[^/]*\.o(?:/|$)".to_vec()),
        );
    }
}