monochromatic-git-policy-cli 0.5.30

//! What:
//!  What happens to the unshown push-failure notices and the unfinished backup pushes of a linked worktree when Git removes
//!  that worktree: before `git worktree remove` or `git worktree prune` runs, the wrapper copies those records into a snapshot
//!  under the common Git directory, and a later wrapper command settles each snapshot, turning it into notices that any
//!  worktree shows when the worktree is gone and discarding it when the removal did not happen.
//! Why:
//!  A worktree keeps its records in its own Git directory, which Git deletes with the worktree. Without this a failed push
//!  nobody had seen yet, or a push the service had not run yet, would vanish with no word, and the person would never learn that a
//!  branch is not on the remote. The wrapper hands the process over to Git for the removal, so it cannot look afterwards; the
//!  snapshot is taken for every linked worktree that has records and settled later. A Git directory that is gone proves the
//!  removal. A Git directory that is still there proves a refusal only once the command that took the snapshot has ended
//!  (the process keeps its identity when it becomes Git, so the snapshot names it) and the worktree's checkout is intact;
//!  while that command lives, or a checkout is missing from a Git directory that is still registered, the removal may be
//!  under way or cut short, and the snapshot is kept. A snapshot that cannot be fully promoted, or fully taken, is kept as far as
//!  it goes and reported on standard error, because a line in the diagnostic log is not a word to the person.
//!
//! In TS you'd write (pseudocode):
//! ```ts
//! const problems = await snapshotLinkedWorktrees(commonDir); // before `git worktree remove`
//! const later = await settleRemovals(commonDir); // on the next command
//! ```

use crate::diagnostic_log::{debug, warn};
use crate::private_storage::{create_private_directory_all, write_private_file_unsynced};
use crate::process_identity::{current_birth_identity, process_birth_identity};
use crate::push_failure_notice::{ORPHAN_NOTICE_DIRECTORY, notice_directory, record_lost_push};
use crate::service_instance::instance_of;
use crate::work_item::{CLAIMED, PENDING, RECOVERING, WORK_DIRECTORY, WorkItem, WorkKind, find_claim, id_of_entry};
use std::path::{Path, PathBuf};

/// Tag used for diagnostics from this module.
const TAG: &str = "worktreeRemovalRescue";

/// Directory under the common Git directory that holds the snapshots.
pub const SNAPSHOT_DIRECTORY: &str = "cli-git/removal-snapshots";

/// File in a snapshot that names the Git directory it was taken from; written last, so a snapshot that has it is whole.
const GIT_DIR_FILE: &str = "gitdir";

/// Directory in a snapshot that holds copies of the notices.
const NOTICES: &str = "notices";

/// Directory in a snapshot that holds copies of the unfinished push records, named `<state>-<record>`.
const WORK: &str = "work";

/// Directory in a snapshot that holds one file per command that took or retook it, named by the command's process number
///  and holding the process's birth identity.
const ATTEMPTS: &str = "attempts";

/// How many rounds of looks a push record gets: `pending`, then the claim. A record moves from the first to the second when the
///  service claims it and back when a claim is given up; the second round catches a record that moved away and back during
///  the first.
const LOOK_ROUNDS: usize = 2;

/// What a Git directory, the process that took a snapshot, and the checkout say about a removal.
#[derive(Clone, Copy, Debug, Eq, PartialEq)]
enum Removal {
    /// The Git directory is gone: Git removed the worktree.
    Done,
    /// Every command that took the snapshot has ended and the checkout is intact: Git refused, or it was a dry run.
    Refused,
    /// A command that took the snapshot may still be removing, the checkout is missing from a registered worktree, or the
    ///  evidence cannot be read: not decided yet.
    Undecided,
}

/// The files of a directory, or none when it cannot be read.
fn files_of(directory: &Path) -> Vec<PathBuf> {
    let Ok(listing) = std::fs::read_dir(directory) else { return Vec::new() };
    let mut paths: Vec<PathBuf> = listing.flatten().map(|entry| { return entry.path(); }).filter(|path| { return path.is_file(); }).collect();
    paths.sort();
    return paths;
}

/// The name of a file as text.
fn name_of(path: &Path) -> String {
    return path.file_name().map(|name| { return name.to_string_lossy().into_owned(); }).unwrap_or_default();
}

/// Writes a file into a snapshot unless a file of that name is already there, which an earlier snapshot of the same worktree
///  left and which holds the same record.
fn save_once(path: &Path, bytes: &[u8]) -> std::io::Result<()> {
    return match write_private_file_unsynced(path, bytes) {
        Err(error) if error.kind() != std::io::ErrorKind::AlreadyExists => Err(error),
        Ok(()) | Err(_) => Ok(()),
    };
}

/// Copies one file into `directory` under `name`; `false` when the file is gone, because whoever shows or finishes it was
///  quicker than the copy and there is nothing left to keep.
fn copy_into(directory: &Path, name: &str, source: &Path) -> std::io::Result<bool> {
    let bytes: Vec<u8> = match std::fs::read(source) {
        Ok(read) => read,
        Err(error) if error.kind() == std::io::ErrorKind::NotFound => return Ok(false),
        Err(error) => return Err(error),
    };
    create_private_directory_all(directory)?;
    save_once(directory.join(name).as_path(), bytes.as_slice())?;
    return Ok(true);
}

/// The ids of the push records of a worktree, in any state a push is unfinished in, once each and in order. A claim's file name
///  carries its generation and clock reading, so the id is the part before them.
fn push_record_names(work_directory: &Path) -> Vec<String> {
    let prefix: String = format!("{}-", WorkKind::Push.name());
    let mut names: Vec<String> = [PENDING, CLAIMED, RECOVERING]
        .into_iter()
        .flat_map(|state| { return files_of(work_directory.join(state).as_path()); })
        .map(|path| { return id_of_entry(name_of(path.as_path()).as_str()).to_owned(); })
        .filter(|name| { return name.starts_with(prefix.as_str()); })
        .collect();
    names.sort();
    names.dedup();
    return names;
}

/// What:
///  The state a push record is in now and its bytes, or nothing when it is in neither state.
/// Why:
///   The service moves a record from `pending` to `claimed` at any moment, and renames a claim each time it renews it. A record
///  listed in one state may be in the other by the time it is read, so the claim is found by its id each time, and a record in
///  neither has finished, so there is nothing to keep. Only a failure to read a record that is there is an error.
///
/// In TS you'd write (pseudocode):
/// ```ts
/// await locatePushRecord(workDirectory, name) // ['claimed', bytes] | undefined
/// ```
fn locate_push_record(work_directory: &Path, name: &str) -> std::io::Result<Option<(&'static str, Vec<u8>)>> {
    for _ in 0..LOOK_ROUNDS {
        for state in [PENDING, CLAIMED] {
            let path: PathBuf = if state == PENDING {
                work_directory.join(PENDING).join(name)
            } else {
                match find_claim(work_directory, name) {
                    Some(found) => found,
                    None => continue,
                }
            };
            match std::fs::read(path) {
                Ok(bytes) => return Ok(Some((state, bytes))),
                Err(error) if error.kind() == std::io::ErrorKind::NotFound => {}
                Err(error) => return Err(error),
            }
        }
    }
    return Ok(None);
}

/// Records, in the snapshot, that this process took it and so may be about to remove the worktree.
fn register_attempt(snapshot: &Path) -> std::io::Result<()> {
    let identity: String = current_birth_identity().map_err(|error| { return std::io::Error::other(error.to_string()); })?;
    let directory: PathBuf = snapshot.join(ATTEMPTS);
    create_private_directory_all(directory.as_path())?;
    // The file is named by this process's number, so no other live process owns it: one left by an earlier process of the
    //  same number holds an identity that is not this process's, and is replaced.
    let own: PathBuf = directory.join(std::process::id().to_string());
    match std::fs::remove_file(own.as_path()) {
        Err(error) if error.kind() != std::io::ErrorKind::NotFound => return Err(error),
        Ok(()) | Err(_) => {}
    }
    return write_private_file_unsynced(own.as_path(), identity.as_bytes());
}

/// What:
///  Whether a command that took the snapshot may still be running, which a process that cannot be told apart from it counts as.
/// Why:
///   A process number alone can be reused, so the birth identity recorded with it decides. A record that cannot be read, or a
///  process that cannot be inspected, is an owner that may be alive: discarding the snapshot on a guess would lose the records.
///
/// In TS you'd write (pseudocode):
/// ```ts
/// await attemptsMayBeRunning(snapshot)
/// ```
fn attempt_may_be_running(snapshot: &Path) -> bool {
    return files_of(snapshot.join(ATTEMPTS).as_path()).iter().any(|path| {
        let Ok(pid) = name_of(path.as_path()).parse::<i64>() else { return false };
        let Ok(recorded) = std::fs::read_to_string(path) else { return true };
        if recorded.is_empty() {
            // Being written.
            return true;
        }
        return match process_birth_identity(pid) {
            Ok(Some(now)) => now == recorded,
            Ok(None) => false,
            Err(error) => {
                warn(TAG, format!("the process {pid} of a worktree removal cannot be inspected: {error}").as_str());
                true
            }
        };
    });
}

/// Whether the checkout a registered worktree's Git directory names is still there; nothing when the Git directory does not say.
fn checkout_present(git_dir: &Path) -> Option<bool> {
    let text: String = std::fs::read_to_string(git_dir.join("gitdir")).ok()?;
    let marker: &Path = Path::new(text.trim_end());
    if marker.as_os_str().is_empty() {
        return None;
    }
    return Some(git_dir.join(marker).exists());
}

/// What:
///  What a snapshot's Git directory, its taker and its checkout show of the removal it was taken for.
/// Why:
///   Git deletes the checkout first and the Git directory last, and refuses before it deletes anything. So a missing Git
///  directory is a removal; an intact checkout whose taker has ended is a refusal; anything else may be a removal in progress
///  or one that was cut short, and deciding on it would either lose the records or keep them for nothing.
///
/// In TS you'd write (pseudocode):
/// ```ts
/// await removalOf(gitDir, snapshot) // 'done' | 'refused' | 'undecided'
/// ```
fn removal_of(git_dir: &Path, snapshot: &Path) -> Removal {
    match std::fs::symlink_metadata(git_dir) {
        Err(error) if error.kind() == std::io::ErrorKind::NotFound => return Removal::Done,
        Err(_) => return Removal::Undecided,
        Ok(_) => {}
    }
    if attempt_may_be_running(snapshot) {
        return Removal::Undecided;
    }
    return if checkout_present(git_dir) == Some(true) { Removal::Refused } else { Removal::Undecided };
}

/// What:
///  Copies the unshown notices and the unfinished push records of each linked worktree into a snapshot of that worktree, and
///  says what could not be copied.
/// Why:
///   Taken before a command that may remove worktrees; a worktree with no records has no snapshot. A snapshot taken before is
///  added to and not replaced: the command that took it may still be removing, and its records are what a copy made now,
///  from a half-removed worktree, could lack. A record that vanished while it was being copied was shown or finished and needs
///  no copy; any other failure is reported to the person, never hidden in the diagnostic log, and what was copied is kept.
///
/// In TS you'd write (pseudocode):
/// ```ts
/// const problems: string[] = await snapshotLinkedWorktrees(commonDir)
/// ```
pub fn snapshot_linked_worktrees(common_dir: &Path) -> Vec<String> {
    let mut problems: Vec<String> = Vec::new();
    let Ok(listing) = std::fs::read_dir(common_dir.join("worktrees")) else { return problems };
    for entry in listing.flatten() {
        let git_dir: PathBuf = entry.path();
        if !git_dir.is_dir() {
            continue;
        }
        let notices: Vec<PathBuf> = files_of(notice_directory(git_dir.as_path()).as_path());
        let work_directory: PathBuf = git_dir.join(WORK_DIRECTORY);
        let pushes: Vec<String> = push_record_names(work_directory.as_path());
        if notices.is_empty() && pushes.is_empty() {
            continue;
        }
        let snapshot: PathBuf = common_dir.join(SNAPSHOT_DIRECTORY).join(instance_of(git_dir.as_path()));
        let mut failures: Vec<String> = Vec::new();
        match create_private_directory_all(snapshot.as_path()) {
            Ok(()) => {
                if let Err(error) = register_attempt(snapshot.as_path()) {
                    failures.push(format!("the command could not be recorded as the one removing it ({error})"));
                }
            }
            Err(error) => failures.push(format!("{} could not be created ({error})", snapshot.display())),
        }
        if failures.is_empty() {
            for path in &notices {
                if let Err(error) = copy_into(snapshot.join(NOTICES).as_path(), name_of(path.as_path()).as_str(), path.as_path()) {
                    failures.push(format!("the notice {} could not be copied ({error})", name_of(path.as_path())));
                }
            }
            for name in &pushes {
                match locate_push_record(work_directory.as_path(), name.as_str()) {
                    Ok(Some((state, bytes))) => {
                        let copied: std::io::Result<()> = create_private_directory_all(snapshot.join(WORK).as_path())
                            .and_then(|()| { return save_once(snapshot.join(WORK).join(format!("{state}-{name}")).as_path(), bytes.as_slice()); });
                        if let Err(error) = copied {
                            failures.push(format!("the push record {name} could not be copied ({error})"));
                        }
                    }
                    Ok(None) => {}
                    Err(error) => failures.push(format!("the push record {name} could not be read ({error})")),
                }
            }
            // Last, so that a snapshot that names its worktree is a whole one.
            if let Err(error) = save_once(snapshot.join(GIT_DIR_FILE).as_path(), git_dir.to_string_lossy().as_bytes()) {
                failures.push(format!("the snapshot could not be completed ({error})"));
            }
        }
        if !failures.is_empty() {
            let said: String = format!(
                "Before this removal, the unshown push failures and unfinished pushes of the worktree {} could not all be saved: {}. \
                 The command goes on, so whatever was not saved is lost if Git removes that worktree; what was saved is kept in {}.",
                git_dir.display(),
                failures.join("; "),
                snapshot.display()
            );
            debug(TAG, said.as_str());
            problems.push(said);
        }
    }
    return problems;
}

/// Moves one saved notice to the notices any worktree shows; a notice another command already moved is not a failure.
fn promote_notice(orphans: &Path, path: &Path) -> std::io::Result<()> {
    create_private_directory_all(orphans)?;
    return match std::fs::rename(path, orphans.join(name_of(path))) {
        Err(error) if error.kind() != std::io::ErrorKind::NotFound => Err(error),
        Ok(()) | Err(_) => Ok(()),
    };
}

/// What:
///  Turns what a snapshot of a removed worktree kept into notices any worktree shows, and removes what it promoted; whether
///  everything was promoted.
/// Why:
///   Each item is removed from the snapshot as soon as it is promoted, so a snapshot kept after a failure promotes only what is
///  left on the next try and a notice already shown is not made again. A failure is reported to the person and the snapshot
///  stays, because the snapshot is then the last copy of that record.
///
/// In TS you'd write (pseudocode):
/// ```ts
/// const complete = await promote(commonDir, snapshot, problems)
/// ```
fn promote(common_dir: &Path, snapshot: &Path, problems: &mut Vec<String>) -> bool {
    let orphans: PathBuf = common_dir.join(ORPHAN_NOTICE_DIRECTORY);
    let mut complete: bool = true;
    for path in files_of(snapshot.join(NOTICES).as_path()) {
        if let Err(error) = promote_notice(orphans.as_path(), path.as_path()) {
            complete = false;
            let said: String = format!(
                "The push failure notice {} of a removed worktree could not be kept for you ({error}). It stays in {}; once that is fixed, any git command shows it.",
                name_of(path.as_path()),
                snapshot.display()
            );
            debug(TAG, said.as_str());
            problems.push(said);
        }
    }
    for path in files_of(snapshot.join(WORK).as_path()) {
        let text: String = match std::fs::read_to_string(path.as_path()) {
            Ok(read) => read,
            // Another command promoted it first.
            Err(error) if error.kind() == std::io::ErrorKind::NotFound => continue,
            Err(error) => {
                complete = false;
                let said: String = format!("The saved record {} of a removed worktree's unfinished push could not be read ({error}); it stays in {}.", name_of(path.as_path()), snapshot.display());
                debug(TAG, said.as_str());
                problems.push(said);
                continue;
            }
        };
        match WorkItem::decode(text.as_str()) {
            Ok(found) => match record_lost_push(common_dir, found.branch_ref.as_str(), found.oid.as_str()) {
                Ok(()) => {
                    // Removed now, so that a try after a failure elsewhere in this snapshot does not record the push again.
                    if let Err(error) = std::fs::remove_file(path.as_path()) {
                        warn(TAG, format!("removing the promoted record {} failed: {error}", path.display()).as_str());
                    }
                }
                Err(reason) => {
                    complete = false;
                    let said: String = format!(
                        "The backup push of {} for commit {} had not finished when its worktree was removed, and noting that failed ({reason}). \
                         The record stays in {}; once that is fixed, any git command shows it.",
                        found.branch_ref,
                        found.oid,
                        snapshot.display()
                    );
                    debug(TAG, said.as_str());
                    problems.push(said);
                }
            },
            Err(reason) => {
                // A record that cannot be read now cannot be read later: it is said once and dropped.
                let said: String = format!("A saved record of an unfinished push of a removed worktree is unreadable ({reason}), so that push is not reported: {}", path.display());
                debug(TAG, said.as_str());
                problems.push(said);
                let _dropped = std::fs::remove_file(path.as_path());
            }
        }
    }
    return complete;
}

/// What:
///  Settles every snapshot whose removal is decided: one whose worktree is still registered and whose taker has ended is
///  discarded; one whose worktree is gone turns its notices into notices any worktree shows, and each unfinished push into a
///  record that says the push may not have finished. Gives what could not be settled, for the person.
/// Why:
///   Whether the removal happened is known only now (`removal_of`). A snapshot whose removal may be under way is left alone,
///  and one that could not be fully promoted is kept for the next try. When nothing was snapshotted this is one directory read.
///
/// In TS you'd write (pseudocode):
/// ```ts
/// const problems: string[] = await settleRemovals(commonDir)
/// ```
pub fn settle_removals(common_dir: &Path) -> Vec<String> {
    let mut problems: Vec<String> = Vec::new();
    let Ok(listing) = std::fs::read_dir(common_dir.join(SNAPSHOT_DIRECTORY)) else { return problems };
    for entry in listing.flatten() {
        let snapshot: PathBuf = entry.path();
        let Ok(named) = std::fs::read_to_string(snapshot.join(GIT_DIR_FILE)) else {
            // A snapshot being written has no name yet; one that never got it is not worth keeping once it is old, but a
            //  concurrent writer may be at it, so it is left to the next settle.
            continue;
        };
        match removal_of(Path::new(named.as_str()), snapshot.as_path()) {
            Removal::Undecided => {}
            Removal::Refused => {
                let _discarded = std::fs::remove_dir_all(snapshot.as_path());
            }
            Removal::Done => {
                if promote(common_dir, snapshot.as_path(), &mut problems) {
                    let _settled = std::fs::remove_dir_all(snapshot.as_path());
                }
            }
        }
    }
    return problems;
}

/// Snapshot controls stay out of the release executable.
#[cfg(test)]
#[path = "worktree_removal_rescue_tests.rs"]
mod tests;