RustProxy/src/admin/file_logs_io.rs

// Bound candidate retention while scanning potentially enormous rotation folders.
// The heap's top is the least desirable result: oldest timestamp, then
// lexicographically last name. Final output preserves newest/name order.
fn newest_log_files(
    candidates: impl Iterator<Item = (String, std::path::PathBuf, u64)>,
    limit: usize,
) -> Vec<(String, std::path::PathBuf, u64)> {
    use std::{cmp::Reverse, collections::BinaryHeap};
    if limit == 0 { return Vec::new(); }
    let mut bounded = BinaryHeap::with_capacity(limit);
    for (name, path, modified_ms) in candidates {
        let item = (Reverse(modified_ms), name, path);
        if bounded.len() < limit {
            bounded.push(item);
        } else if bounded.peek().is_some_and(|worst| item.cmp(worst).is_lt()) {
            bounded.pop();
            bounded.push(item);
        }
    }
    let mut result = bounded.into_iter()
        .map(|(Reverse(modified), name, path)| (name, path, modified))
        .collect::<Vec<_>>();
    result.sort_by(|a, b| b.2.cmp(&a.2).then_with(|| a.0.cmp(&b.0)));
    result
}

// Bounded file-log inventory helpers and gzip tail reader.
fn rotated_gzip_sources(
    active: &std::path::Path,
    id_prefix: &str,
    label_prefix: &str,
    kind: &'static str,
    enabled: bool,
    format: &'static str,
) -> Vec<FileLogSource> {
    let Some(parent) = active.parent() else { return Vec::new(); };
    let Some(base) = active.file_name().and_then(|name| name.to_str()) else { return Vec::new(); };
    let safe = std::fs::symlink_metadata(parent).ok().is_some_and(|meta| meta.file_type().is_dir() && !meta.file_type().is_symlink());
    if !safe { return Vec::new(); }
    let prefix = format!("{base}.");
    let entries = newest_log_files(std::fs::read_dir(parent).ok().into_iter().flat_map(|iter| iter.filter_map(|entry| entry.ok())).filter_map(|entry| {
        let name = entry.file_name().to_string_lossy().into_owned();
        if name.len() > 180 || !name.starts_with(&prefix) || !name.ends_with(".gz") { return None; }
        let file_type = entry.file_type().ok()?;
        if !file_type.is_file() || file_type.is_symlink() { return None; }
        let meta = entry.metadata().ok()?;
        Some((name, entry.path(), file_modified_ms(&meta)))
    }), 64);
    entries.into_iter().map(|(name,path,_)| source_from_path(
        format!("{id_prefix}:gz:{name}"),
        format!("{label_prefix} · {name}"),
        id_prefix.to_owned(),
        kind, path, enabled, format
    )).collect()
}

fn open_file_log(path: &std::path::Path) -> std::io::Result<std::fs::File> {
    let mut options = std::fs::OpenOptions::new();
    options.read(true);
    #[cfg(unix)]
    {
        use std::os::unix::fs::OpenOptionsExt;
        options.custom_flags(libc::O_NOFOLLOW);
    }
    let file = options.open(path)?;
    let metadata = file.metadata()?;
    if !metadata.file_type().is_file() {
        return Err(std::io::Error::new(
            std::io::ErrorKind::InvalidInput,
            "configured log path is not a regular file",
        ));
    }
    Ok(file)
}
fn read_tail(
    source: &FileLogSource,
    budget: u64,
    line_budget: usize,
) -> std::io::Result<(Vec<String>, u64, bool)> {
    use std::io::{Read, Seek, SeekFrom};
    let mut file = open_file_log(&source.path)?;
    let length = file.metadata()?.len();
    if source.compressed {
        // gzip streams are not seekable. Read a complete bounded rotation, then
        // keep only the same 2 MiB decompressed tail used for plain files.
        if length > budget || length > FILE_LOG_GZIP_INPUT_MAX {
            return Ok((Vec::new(), length.min(budget), true));
        }
        let mut decoder = flate2::read::GzDecoder::new(file);
        let mut decoded = Vec::new();
        decoder
            .by_ref()
            .take(FILE_LOG_GZIP_OUTPUT_MAX + 1)
            .read_to_end(&mut decoded)?;
        if decoded.len() as u64 > FILE_LOG_GZIP_OUTPUT_MAX {
            return Ok((Vec::new(), length, true));
        }
        let keep = decoded.len().min(FILE_LOG_SCAN_BYTES_PER_SOURCE as usize);
        let start = decoded.len().saturating_sub(keep);
        let mut content = decoded[start..].to_vec();
        let truncated = start > 0;
        if truncated {
            if let Some(index) = content.iter().position(|byte| *byte == b'\n') {
                content.drain(..=index);
            } else {
                content.clear();
            }
        }
        let text = String::from_utf8_lossy(&content);
        let mut iter = text.lines().rev();
        let lines = iter.by_ref().take(line_budget).map(str::to_owned).collect::<Vec<_>>();
        return Ok((lines, length, truncated || iter.next().is_some()));
    }
    let bytes = length.min(budget).min(FILE_LOG_SCAN_BYTES_PER_SOURCE);
    let start = length.saturating_sub(bytes);
    file.seek(SeekFrom::Start(start))?;
    let mut content = Vec::with_capacity(bytes as usize);
    file.take(bytes).read_to_end(&mut content)?;
    let truncated = start > 0;
    if truncated {
        if let Some(index) = content.iter().position(|byte| *byte == b'\n') {
            content.drain(..=index);
        } else {
            content.clear();
        }
    }
    let text = String::from_utf8_lossy(&content);
    let mut iter = text.lines().rev();
    let lines = iter.by_ref().take(line_budget).map(str::to_owned).collect::<Vec<_>>();
    Ok((lines, bytes, truncated || iter.next().is_some()))
}