Artifact 59514281b8bd424e3634a96afafd74f8bf0e0395977fa30562422ac9a2144da2:
- File
internal/checkout/extract.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: 12215) [annotate] [blame] [check-ins using]
package checkout import ( "context" "fmt" "os" "path/filepath" "strconv" "strings" "github.com/danmestas/go-libfossil/internal/content" libfossil "github.com/danmestas/go-libfossil/internal/fsltype" "github.com/danmestas/go-libfossil/internal/hash" "github.com/danmestas/go-libfossil/internal/manifest" "github.com/danmestas/go-libfossil/internal/vfile" ) // extractSingleFile expands a blob and writes it to the checkout directory, // returning the vfile.mtime to record for it (writtenMTime). // Skips writing if dryRun is true. func (c *Checkout) extractSingleFile( pathname string, blobRid int64, isexe int64, dryRun bool, ) (int64, error) { if dryRun { return 0, nil } data, err := content.Expand(c.repo.DB(), libfossil.FslID(blobRid)) if err != nil { return 0, fmt.Errorf("checkout.Extract: expand blob for %s: %w", pathname, err) } fullPath, err := c.safePath(pathname) if err != nil { return 0, fmt.Errorf("checkout.Extract: path traversal in %s: %w", pathname, err) } parentDir := filepath.Dir(fullPath) if err := c.env.Storage.MkdirAll(parentDir, 0o755); err != nil { return 0, fmt.Errorf("checkout.Extract: mkdir %s: %w", parentDir, err) } perm := os.FileMode(0o644) if isexe != 0 { perm = 0o755 } if err := c.env.Storage.WriteFile(fullPath, data, perm); err != nil { return 0, fmt.Errorf("checkout.Extract: write %s: %w", fullPath, err) } mtime, err := c.writtenMTime(fullPath) if err != nil { return 0, fmt.Errorf("checkout.Extract: %w", err) } return mtime, nil } // Extract writes files from the specified checkin to disk via simio.Storage. // Populates vfile, updates vvar checkout/checkout-hash to rid. // // If opts.DryRun is true, skips writing files but still calls observer and callback. // If opts.Force is false (default), refuses before changing anything when the // current version has unsaved changes or an untracked file would be // overwritten (see checkSafeToExtract). // // Panics if c is nil (TigerStyle precondition). func (c *Checkout) Extract(rid libfossil.FslID, opts ExtractOpts) error { if c == nil { panic("checkout.Extract: nil *Checkout") } // Start observer ctx := c.obs.ExtractStarted(context.Background(), ExtractStart{ Operation: "extract", TargetRID: rid, }) var filesWritten int var extractErr error defer func() { c.obs.ExtractCompleted(ctx, ExtractEnd{ Operation: "extract", TargetRID: rid, FilesWritten: filesWritten, Err: extractErr, }) }() // Check before LoadVFile: it replaces the current version's rows, and a // refusal must leave the checkout exactly as it was. if !opts.Force && !opts.DryRun { if err := c.checkSafeToExtract(rid); err != nil { extractErr = err return extractErr } } // The files come from the target's manifest, so a dry run never has to // load them into vfile; only a real extract replaces the file list. vfRows, err := c.extractRows(rid) if err != nil { extractErr = err return extractErr } if !opts.DryRun { if err := c.replaceFileList(rid, opts.Force); err != nil { extractErr = err return extractErr } } for _, row := range vfRows { if err := c.extractRow(ctx, rid, row, opts); err != nil { extractErr = err return extractErr } filesWritten++ } if opts.DryRun { return nil } if extractErr = c.finalizeExtract(rid, opts.Force); extractErr != nil { return extractErr } if opts.SetMTime { extractErr = c.setFileMTimes(rid) } return extractErr } // finalizeExtract looks up the blob UUID for rid and updates the vvar // checkout/checkout-hash entries. // It also ends a pending merge on a switch to another version or a forced // extract, but not on a plain re-extract of the version already checked out, // which fossil's checkout does not treat as a switch. func (c *Checkout) finalizeExtract(rid libfossil.FslID, force bool) error { prior, _, err := c.Version() if err != nil { return fmt.Errorf("checkout.Extract: %w", err) } endMerge := force || prior != rid var uuid string err = c.repo.DB().QueryRow("SELECT uuid FROM blob WHERE rid = ?", int64(rid)).Scan(&uuid) if err != nil { return fmt.Errorf("checkout.Extract: query blob uuid: %w", err) } if err := setVVar(c.db, "checkout", strconv.FormatInt(int64(rid), 10)); err != nil { return fmt.Errorf("checkout.Extract: %w", err) } if err := setVVar(c.db, "checkout-hash", uuid); err != nil { return fmt.Errorf("checkout.Extract: %w", err) } if !endMerge { return nil } // A switch ends any pending merge, as fossil's checkout does; a stale // vmerge row would give the next commit a bogus merge parent. if _, err := c.db.Exec("DELETE FROM vmerge"); err != nil { return fmt.Errorf("checkout.Extract: clear vmerge: %w", err) } return nil } // replaceFileList loads version rid into vfile. A forced extract first drops // every row, the current version's included, as fossil's checkout --force // does, so no pending add, rename or removal survives it; otherwise rows the // version already has are kept (LoadVFile). func (c *Checkout) replaceFileList(rid libfossil.FslID, force bool) error { if rid <= 0 { panic("checkout.replaceFileList: rid must be positive") } if force { if _, err := c.db.Exec("DELETE FROM vfile"); err != nil { return fmt.Errorf("checkout.Extract: clear vfile: %w", err) } } if err := c.LoadVFile(rid, true); err != nil { return fmt.Errorf("checkout.Extract: %w", err) } return nil } // extractRow is one vfile row Extract writes to disk. type extractRow struct { pathname string blobRid int64 isexe int64 } // extractRows lists version rid's files with the blob each one's content is // in, from its manifest. It refuses a version whose content is missing // (resolveFiles), before anything is written. func (c *Checkout) extractRows(rid libfossil.FslID) ([]extractRow, error) { if rid <= 0 { panic("checkout.extractRows: rid must be positive") } files, blobRIDs, err := c.resolveFiles(rid) if err != nil { return nil, fmt.Errorf("checkout.Extract: %w", err) } out := make([]extractRow, len(files)) for i, f := range files { isexe := int64(0) if strings.Contains(f.Perm, "x") { isexe = 1 } out[i] = extractRow{pathname: f.Name, blobRid: int64(blobRIDs[i]), isexe: isexe} } return out, nil } // extractRow writes one file of version vid, records the mtime the write // gave it, and reports it to the observer and the caller's callback. func (c *Checkout) extractRow( ctx context.Context, vid libfossil.FslID, row extractRow, opts ExtractOpts, ) error { mtime, err := c.extractSingleFile(row.pathname, row.blobRid, row.isexe, opts.DryRun) if err != nil { return err } if !opts.DryRun { if err := c.recordMTime(vid, row.pathname, row.blobRid, mtime); err != nil { return fmt.Errorf("checkout.Extract: %w", err) } } c.obs.ExtractFileCompleted(ctx, row.pathname, UpdateAdded) if opts.Callback != nil { if err := opts.Callback(row.pathname, UpdateAdded); err != nil { return fmt.Errorf("checkout.Extract: callback for %s: %w", row.pathname, err) } } return nil } // checkSafeToExtract refuses an Extract that would destroy work the user has // not committed. It follows fossil's checkout command, which refuses while the // current version has unsaved changes (edits, adds, removals, renames, // merges) and prompts before overwriting a file it does not track; a library // cannot prompt, so it refuses. It runs before anything in the checkout // changes. // // One relaxation: an edit whose content already equals what target would // write loses nothing, so it does not block. That matters because Create // records the tip as checked out before any file is written, so a directory // already holding an older version looks edited relative to the tip. func (c *Checkout) checkSafeToExtract(target libfossil.FslID) error { if target <= 0 { panic("checkout.checkSafeToExtract: target must be positive") } current, _, err := c.Version() if err != nil { return fmt.Errorf("checkout.Extract: %w", err) } files, err := manifest.ListFiles(c.repo, target) if err != nil { return fmt.Errorf("checkout.Extract: %w", err) } targetHash := make(map[string]string, len(files)) for _, f := range files { targetHash[f.Name] = f.UUID } if current > 0 { atRisk, err := c.firstChangeAtRisk(current, targetHash) if err != nil { return fmt.Errorf("checkout.Extract: %w", err) } if atRisk != "" { return fmt.Errorf( "checkout.Extract: file %s has local changes; use Force to overwrite", atRisk, ) } } return c.checkNoUntrackedClobber(current, files) } // firstChangeAtRisk scans version vid and returns the first changed file // whose change extracting target would lose, or "" when there is none. func (c *Checkout) firstChangeAtRisk( vid libfossil.FslID, targetHash map[string]string, ) (string, error) { if vid <= 0 { panic("checkout.firstChangeAtRisk: vid must be positive") } if targetHash == nil { panic("checkout.firstChangeAtRisk: nil targetHash") } if err := c.refreshChanged(vid); err != nil { return "", err } // A pending merge is unsaved work in its own right: the vmerge record // would be dropped even if every merged file already matches target. var merging bool if err := c.db.QueryRow( "SELECT EXISTS(SELECT 1 FROM vmerge)", ).Scan(&merging); err != nil { return "", fmt.Errorf("query vmerge: %w", err) } rows, err := vfile.Load(c.db, int64(vid)) if err != nil { return "", err } for _, r := range rows { change := vfile.Classify(r, false) if change == ChangeNone { continue } // Only a plain edit (no merge pending, not a merge state) of a file // target also writes can be safe; its content decides. if !merging && change == ChangeModified && !r.PendingMerge() && !r.ModeChanged() { if uuid, ok := targetHash[r.Pathname]; ok { clobber, err := c.wouldClobber(r.Pathname, uuid) if err != nil { return "", err } if !clobber { continue } } } return r.Pathname, nil } return "", nil } // wouldClobber reports whether writing the artifact uuid to pathname would // destroy content: true only when a file is there and differs from it. func (c *Checkout) wouldClobber(pathname, uuid string) (bool, error) { if !hash.IsValidHash(uuid) { panic("checkout.wouldClobber: invalid uuid for " + pathname) } fullPath, err := c.safePath(pathname) if err != nil { return false, nil // Extraction rejects the path itself, with a better error. } data, err := c.env.Storage.ReadFile(fullPath) if err != nil { if os.IsNotExist(err) { return false, nil } return false, fmt.Errorf("checkout.Extract: read %s: %w", fullPath, err) } return hash.ContentHash(data, uuid) != uuid, nil } // checkNoUntrackedClobber returns an error if a file that version current // does not track exists on disk where target would write different content. // current may be 0 (nothing checked out yet). func (c *Checkout) checkNoUntrackedClobber( current libfossil.FslID, files []manifest.FileEntry, ) error { if current < 0 { panic("checkout.checkNoUntrackedClobber: negative current") } tracked, err := c.trackedPaths(current) if err != nil { return err } for _, f := range files { if tracked[f.Name] { continue } clobber, err := c.wouldClobber(f.Name, f.UUID) if err != nil { return err } if clobber { return fmt.Errorf( "checkout.Extract: untracked file %s has local changes; use Force to overwrite", f.Name, ) } } return nil } // trackedPaths returns the set of pathnames vfile holds for version vid. func (c *Checkout) trackedPaths(vid libfossil.FslID) (map[string]bool, error) { rows, err := c.db.Query("SELECT pathname FROM vfile WHERE vid = ?", int64(vid)) if err != nil { return nil, fmt.Errorf("checkout.Extract: query vfile: %w", err) } defer rows.Close() tracked := make(map[string]bool) for rows.Next() { var name string if err := rows.Scan(&name); err != nil { return nil, fmt.Errorf("checkout.Extract: scan vfile: %w", err) } tracked[name] = true } if err := rows.Err(); err != nil { return nil, fmt.Errorf("checkout.Extract: iterate vfile: %w", err) } return tracked, nil }