Artifact [f9e5e2d88a]

Artifact f9e5e2d88a0d19e0228161819fff6fabea246b3d4e013d641429a2b9645983b2:

• File internal/checkout/update.go — part of check-in [a71bf662ee] at 2026-10-18 20:59:23 on branch main — fix(checkout): record and trust vfile.mtime as fossil does (#444)

  * fix(checkout): record and trust vfile.mtime as fossil does

  go-libfossil wrote vfile.mtime = 0 everywhere and hashed every file on every scan, so it and fossil rehashed a go-libfossil checkout on every command.

  The scan is now fossil's vfile_check_signature: a file whose size matches its artifact (blob.size of mrid) and whose mtime matches vfile.mtime is taken as unchanged without being read, unless ScanHash is given or the mtime-changes setting is off; the scan stores the file's mtime in whole seconds and runs its updates in one transaction. Status, commit, merge, revert and extract use the default scan, as fossil's commands do.

  Mtimes are recorded where fossil records them: extract and update store the mtime of each file written with its artifact's content, an unchanged file keeps its mtime through LoadVFile (fossil's update transfer and in-place commit), committed files keep theirs, revert stores the restored file's, and SetMTime stores the time it gives. Undoing a rename drops the mtime, since the file was not checked.

  One deliberate difference: an mtime not older than the current second is stored as 0, so the next scan (ours or fossil's) hashes the file and an edit that keeps the size within that second is still seen. A stored 0 is never trusted.

  MemStorage now stamps each write with a new mtime from its own clock, as a filesystem does, so in-memory checkouts see same-size edits.

  Status on the sqlite repository's 2,220-file checkout: about 250 ms before, about 30 ms after.

  * test(checkout): pin the same-second guard against fossil

  A same-size edit in the second a file was scanned is committed, and fossil 2.28 reading a row the guard stored as 0 reports the edit. Both tests fail with the guard removed. The guard's comment and the CHANGELOG name it as the one deliberate difference from vfile_check_signature.

  * fix(checkout): record the mtime of each write as it happens

  Extract and update statted their files only after every file was written and vfile reloaded, so an edit in between was recorded as the written content. They now stat each file right after writing it, as fossil's vfile_to_disk does, and carry that mtime to the row. Revert and the new writtenMTime read the clock before the stat, as the scan does.

  A caller's Env without a Clock now gets the real clock, as Storage defaults to the OS; scans read it. The mtime-changes setting is read from the repository only: go-libfossil reads no global fossil settings, which the comment and CHANGELOG now say.

  New tests: an edit after extract's or update's write is seen, a stored 0 is never trusted, and a simulated clock at a realistic time drives the mtime fast path. (user: [email protected], size: 14423) [annotate] [blame] [check-ins using]

package checkout

import (
	"context"
	"fmt"
	"os"
	"path/filepath"
	"sort"
	"strconv"

	"github.com/danmestas/go-libfossil/internal/content"
	libfossil "github.com/danmestas/go-libfossil/internal/fsltype"
	"github.com/danmestas/go-libfossil/internal/manifest"
	"github.com/danmestas/go-libfossil/internal/merge"
)

// updateFileMaps holds name→UUID maps for the current, target, and ancestor versions.
type updateFileMaps struct {
	current  map[string]string
	target   map[string]string
	ancestor map[string]string
	allNames map[string]bool
}

// buildFileMaps builds name→UUID maps for the current, target, and ancestor
// file lists, plus the union of all file names.
func buildFileMaps(currentFiles, targetFiles, ancestorFiles []manifest.FileEntry) updateFileMaps {
	m := updateFileMaps{
		current:  make(map[string]string, len(currentFiles)),
		target:   make(map[string]string, len(targetFiles)),
		ancestor: make(map[string]string, len(ancestorFiles)),
		allNames: make(map[string]bool, len(currentFiles)+len(targetFiles)),
	}
	for _, f := range currentFiles {
		m.current[f.Name] = f.UUID
		m.allNames[f.Name] = true
	}
	for _, f := range targetFiles {
		m.target[f.Name] = f.UUID
		m.allNames[f.Name] = true
	}
	for _, f := range ancestorFiles {
		m.ancestor[f.Name] = f.UUID
	}
	return m
}

// processFileUpdates iterates over all file names and applies update/merge
// logic for each one. Returns the paths actually written, removed, and
// conflicted — not merely counts — so the caller can report exactly which
// files a consumer needs to inspect. Each file written with the target's
// content as it is (not merged) goes into fresh, with the vfile.mtime its
// write gave it (writtenMTime).
//
// Bounded: the loop and the returned slices are each bounded by
// len(maps.allNames), the union of the current and target manifests for
// this update — finite for any real checkin.
func (c *Checkout) processFileUpdates(
	ctx context.Context,
	maps updateFileMaps,
	strategy merge.Strategy,
	opts UpdateOpts,
	fresh map[string]int64,
) (filesWritten, filesRemoved, conflicted []string, err error) {
	for name := range maps.allNames {
		curUUID, inCurrent := maps.current[name]
		tgtUUID, inTarget := maps.target[name]
		ancUUID := maps.ancestor[name]

		change, ferr := c.updateFile(
			name, curUUID, tgtUUID, ancUUID,
			inCurrent, inTarget, strategy, opts.DryRun, fresh,
		)
		if ferr != nil {
			return filesWritten, filesRemoved, conflicted, fmt.Errorf("checkout.Update: %w", ferr)
		}

		if change == UpdateNone {
			continue
		}

		switch change {
		case UpdateAdded, UpdateUpdated, UpdateMerged:
			filesWritten = append(filesWritten, name)
		case UpdateConflictMerged:
			filesWritten = append(filesWritten, name)
			conflicted = append(conflicted, name)
			c.obs.Error(ctx, fmt.Errorf(
				"checkout.Update: merge conflict in %s", name,
			))
		case UpdateRemoved:
			filesRemoved = append(filesRemoved, name)
		}

		c.obs.ExtractFileCompleted(ctx, name, change)

		if opts.Callback != nil {
			if cerr := opts.Callback(name, change); cerr != nil {
				return filesWritten, filesRemoved, conflicted, fmt.Errorf(
					"checkout.Update: callback for %s: %w", name, cerr,
				)
			}
		}
	}
	return filesWritten, filesRemoved, conflicted, nil
}

// buildUpdateMaps gathers file lists for current, target, and their common
// ancestor, then builds the combined name→UUID maps for 3-way comparison.
func (c *Checkout) buildUpdateMaps(currentRID, target libfossil.FslID) (updateFileMaps, error) {
	currentFiles, err := manifest.ListFiles(c.repo, currentRID)
	if err != nil {
		return updateFileMaps{}, fmt.Errorf("checkout.Update: list current files: %w", err)
	}
	targetFiles, err := manifest.ListFiles(c.repo, target)
	if err != nil {
		return updateFileMaps{}, fmt.Errorf("checkout.Update: list target files: %w", err)
	}

	ancestor, err := merge.FindCommonAncestor(c.repo, currentRID, target)
	if err != nil {
		ancestor = currentRID
	}

	var ancestorFiles []manifest.FileEntry
	if ancestor == currentRID {
		ancestorFiles = currentFiles
	} else {
		ancestorFiles, err = manifest.ListFiles(c.repo, ancestor)
		if err != nil {
			return updateFileMaps{}, fmt.Errorf("checkout.Update: list ancestor files: %w", err)
		}
	}

	return buildFileMaps(currentFiles, targetFiles, ancestorFiles), nil
}

// Update updates the checkout to a new version, performing 3-way merge where
// needed to preserve local modifications.
//
// If opts.TargetRID is 0, CalcUpdateVersion is used to find the target.
// If the target equals the current version, Update returns a zero-value
// UpdateResult and a nil error (nothing to do).
//
// A non-nil error means the update genuinely failed; the returned
// UpdateResult's contents are not meaningful in that case (see UpdateResult
// for the outcome mapping). Update never returns a non-nil error together
// with a non-empty UpdateResult.Conflicted — a conflicted merge is a
// success, not a failure.
//
// Panics if c is nil (TigerStyle precondition).
func (c *Checkout) Update(opts UpdateOpts) (UpdateResult, error) {
	if c == nil {
		panic("checkout.Update: nil *Checkout")
	}

	target, currentRID, err := c.updateTarget(opts.TargetRID)
	if err != nil {
		return UpdateResult{}, err
	}
	if target == 0 {
		return UpdateResult{}, nil // nothing to update, or already there
	}

	ctx := c.obs.ExtractStarted(context.Background(), ExtractStart{
		Operation: "update",
		TargetRID: target,
	})

	var filesWritten, filesRemoved, conflicted []string
	fresh := map[string]int64{}
	var updateErr error
	defer func() {
		c.obs.ExtractCompleted(ctx, ExtractEnd{
			Operation:    "update",
			TargetRID:    target,
			FilesWritten: len(filesWritten),
			FilesRemoved: len(filesRemoved),
			Conflicts:    len(conflicted),
			Err:          updateErr,
		})
	}()

	// Refuse a target whose content is not all here before writing any file,
	// as fossil's update does; finalizeUpdate would refuse it too, but only
	// after disk had moved.
	if _, _, err := c.resolveFiles(target); err != nil {
		updateErr = fmt.Errorf("checkout.Update: %w", err)
		return UpdateResult{}, updateErr
	}

	maps, err := c.buildUpdateMaps(currentRID, target)
	if err != nil {
		updateErr = err
		return UpdateResult{}, updateErr
	}

	strategy, ok := merge.StrategyByName("three-way")
	if !ok {
		updateErr = fmt.Errorf("checkout.Update: three-way merge strategy not registered")
		return UpdateResult{}, updateErr
	}

	filesWritten, filesRemoved, conflicted, updateErr = c.processFileUpdates(ctx, maps, strategy, opts, fresh)
	if updateErr != nil {
		return UpdateResult{}, updateErr
	}

	// Finalize: reload vfile and update vvar
	if !opts.DryRun {
		if updateErr = c.finalizeUpdate(target, fresh); updateErr != nil {
			return UpdateResult{}, updateErr
		}
		if opts.SetMTime {
			if updateErr = c.setFileMTimes(target); updateErr != nil {
				return UpdateResult{}, updateErr
			}
		}
	}

	return sortedUpdateResult(filesWritten, filesRemoved, conflicted), nil
}

// sortedUpdateResult builds Update's result with each path list sorted.
// processFileUpdates walks a map, so the order the lists were built in is
// randomized by the Go runtime; sorting makes the result deterministic, so
// golden-file and reflect.DeepEqual assertions against it are meaningful.
func sortedUpdateResult(written, removed, conflicted []string) UpdateResult {
	sort.Strings(written)
	sort.Strings(removed)
	sort.Strings(conflicted)
	return UpdateResult{
		FilesWritten: written,
		FilesRemoved: removed,
		Conflicted:   conflicted,
	}
}

// updateTarget picks the version Update moves to and the current one. A
// target of 0 means there is nothing to do: no newer version, or the
// checkout is already there.
func (c *Checkout) updateTarget(
	requested libfossil.FslID,
) (target, current libfossil.FslID, err error) {
	if c == nil {
		panic("checkout.updateTarget: nil *Checkout")
	}
	if requested < 0 {
		return 0, 0, fmt.Errorf("checkout.Update: invalid target rid %d", requested)
	}

	target = requested
	if target == 0 {
		target, err = c.CalcUpdateVersion()
		if err != nil {
			return 0, 0, fmt.Errorf("checkout.Update: %w", err)
		}
		if target == 0 {
			return 0, 0, nil
		}
	}
	current, _, err = c.Version()
	if err != nil {
		return 0, 0, fmt.Errorf("checkout.Update: %w", err)
	}
	if current == target {
		return 0, current, nil
	}
	if target <= 0 {
		panic("checkout.updateTarget: chose a non-positive target")
	}
	return target, current, nil
}

// finalizeUpdate reloads vfile for the target version, records the mtimes
// of the fresh files (written with the target's content, by pathname), looks
// up its UUID,
// and updates the vvar checkout/checkout-hash entries. An unchanged file
// keeps its mtime through LoadVFile, as fossil's update transfers it.
func (c *Checkout) finalizeUpdate(target libfossil.FslID, fresh map[string]int64) error {
	if err := c.LoadVFile(target, true); err != nil {
		return fmt.Errorf("checkout.Update: reload vfile: %w", err)
	}
	for name, mtime := range fresh {
		var rid int64
		if err := c.db.QueryRow(
			"SELECT mrid FROM vfile WHERE vid = ? AND pathname = ?", int64(target), name,
		).Scan(&rid); err != nil {
			return fmt.Errorf("checkout.Update: vfile row for %s: %w", name, err)
		}
		if err := c.recordMTime(target, name, rid, mtime); err != nil {
			return fmt.Errorf("checkout.Update: %w", err)
		}
	}

	var targetUUID string
	err := c.repo.DB().QueryRow("SELECT uuid FROM blob WHERE rid = ?", int64(target)).Scan(&targetUUID)
	if err != nil {
		return fmt.Errorf("checkout.Update: query target uuid: %w", err)
	}

	if err := setVVar(c.db, "checkout", strconv.FormatInt(int64(target), 10)); err != nil {
		return fmt.Errorf("checkout.Update: %w", err)
	}
	if err := setVVar(c.db, "checkout-hash", targetUUID); err != nil {
		return fmt.Errorf("checkout.Update: %w", err)
	}
	return nil
}

// updateFile handles the update logic for a single file across three versions.
// Returns the UpdateChange type applied, or UpdateNone if the file was skipped.
func (c *Checkout) updateFile(
	name string,
	curUUID, tgtUUID, ancUUID string,
	inCurrent, inTarget bool,
	strategy merge.Strategy,
	dryRun bool,
	fresh map[string]int64,
) (UpdateChange, error) {
	switch {
	case !inCurrent && inTarget:
		// New file in target — extract and write
		if !dryRun {
			mtime, err := c.writeFileFromUUID(name, tgtUUID)
			if err != nil {
				return UpdateNone, fmt.Errorf("add %s: %w", name, err)
			}
			fresh[name] = mtime
		}
		return UpdateAdded, nil

	case inCurrent && !inTarget:
		// File removed in target — delete from disk
		if !dryRun {
			fullPath, err := c.safePath(name)
			if err != nil {
				return UpdateNone, fmt.Errorf("remove %s: %w", name, err)
			}
			if err := c.env.Storage.Remove(fullPath); err != nil {
				// Ignore errors if file already missing
				if !os.IsNotExist(err) {
					return UpdateNone, fmt.Errorf("remove %s: %w", name, err)
				}
			}
		}
		return UpdateRemoved, nil

	case inCurrent && inTarget:
		if curUUID == tgtUUID {
			// Same content — skip
			return UpdateNone, nil
		}

		// Different content — determine who changed
		if curUUID == ancUUID {
			// Current matches ancestor, target changed — fast-forward
			if !dryRun {
				mtime, err := c.writeFileFromUUID(name, tgtUUID)
				if err != nil {
					return UpdateNone, fmt.Errorf("update %s: %w", name, err)
				}
				fresh[name] = mtime
			}
			return UpdateUpdated, nil
		}

		if tgtUUID == ancUUID {
			// Target matches ancestor, current changed — keep current
			return UpdateNone, nil
		}

		// Both changed — 3-way merge needed
		return c.mergeFile(name, ancUUID, curUUID, tgtUUID, strategy, dryRun)

	default:
		// Not in current, not in target — impossible given allNames construction
		return UpdateNone, nil
	}
}

// mergeFile performs a 3-way merge of a file that was modified in both current and target.
func (c *Checkout) mergeFile(
	name, ancUUID, curUUID, tgtUUID string,
	strategy merge.Strategy,
	dryRun bool,
) (UpdateChange, error) {
	// Expand all three versions
	ancData, err := c.expandUUID(ancUUID)
	if err != nil {
		return UpdateNone, fmt.Errorf("merge %s: expand ancestor: %w", name, err)
	}
	curData, err := c.expandUUID(curUUID)
	if err != nil {
		return UpdateNone, fmt.Errorf("merge %s: expand current: %w", name, err)
	}
	tgtData, err := c.expandUUID(tgtUUID)
	if err != nil {
		return UpdateNone, fmt.Errorf("merge %s: expand target: %w", name, err)
	}

	result, err := strategy.Merge(ancData, curData, tgtData)
	if err != nil {
		return UpdateNone, fmt.Errorf("merge %s: %w", name, err)
	}

	if !dryRun {
		fullPath, err := c.safePath(name)
		if err != nil {
			return UpdateNone, fmt.Errorf("merge %s: path traversal: %w", name, err)
		}
		parentDir := filepath.Dir(fullPath)
		if err := c.env.Storage.MkdirAll(parentDir, os.FileMode(0o755)); err != nil {
			return UpdateNone, fmt.Errorf("merge %s: mkdir: %w", name, err)
		}
		if err := c.env.Storage.WriteFile(fullPath, result.Content, os.FileMode(0o644)); err != nil {
			return UpdateNone, fmt.Errorf("merge %s: write: %w", name, err)
		}
	}

	if result.Clean {
		return UpdateMerged, nil
	}
	return UpdateConflictMerged, nil
}

// writeFileFromUUID expands a blob by UUID and writes it to the checkout
// directory, returning the vfile.mtime to record for it (writtenMTime).
func (c *Checkout) writeFileFromUUID(name, uuid string) (int64, error) {
	data, err := c.expandUUID(uuid)
	if err != nil {
		return 0, fmt.Errorf("writeFileFromUUID %s: %w", name, err)
	}

	fullPath, err := c.safePath(name)
	if err != nil {
		return 0, fmt.Errorf("writeFileFromUUID %s: path traversal: %w", name, err)
	}
	parentDir := filepath.Dir(fullPath)
	if err := c.env.Storage.MkdirAll(parentDir, os.FileMode(0o755)); err != nil {
		return 0, fmt.Errorf("mkdir %s: %w", parentDir, err)
	}
	if err := c.env.Storage.WriteFile(fullPath, data, os.FileMode(0o644)); err != nil {
		return 0, fmt.Errorf("write %s: %w", fullPath, err)
	}
	return c.writtenMTime(fullPath)
}

// expandUUID resolves a UUID to its expanded blob content.
func (c *Checkout) expandUUID(uuid string) ([]byte, error) {
	rid, ok := content.AvailableByUUID(c.repo.DB(), uuid)
	if !ok {
		return nil, fmt.Errorf("blob not found for uuid %s", uuid)
	}
	data, err := content.Expand(c.repo.DB(), rid)
	if err != nil {
		return nil, fmt.Errorf("expand rid %d (uuid %s): %w", rid, uuid, err)
	}
	return data, nil
}