fix(engine): cross-backend integrity guard with retry-then-damaged

A truncated debrid download (in-memory byte counter hit 100% while the
NFS write-back silently dropped most of the bytes) was marked completed.
The 1.1.6 fsync fix closed the debrid-specific hole; this generalizes the
guarantee so "completed" never means a corrupt file on ANY backend.

- IntegrityError + bounded retry: on a corrupt/short result the manager
  re-downloads the same source up to 3x (clean start), then surfaces the
  task as damaged ("corrupt download:" prefix) instead of completing it.
- verify (size mismatch / empty), debrid (incomplete / post-write / flush),
  torrent (BytesMissing), usenet (par2 unrepairable / repair-failed) all
  classify integrity failures so they route through the retry/damaged path.
- scanner: a file ffprobe can't read is emitted as a damaged library_item
  (reason "unreadable") instead of being silently dropped from the sync.
- tests: manager retry-then-success + retry-exhausted-then-damaged,
  verifying->resolving transition, damaged sync item.
This commit is contained in:
Deivid Soto 2026-06-17 12:51:47 +02:00
parent 271413e0f9
commit a5f3f0914a
13 changed files with 400 additions and 91 deletions

View file

@ -0,0 +1,41 @@
package engine
import (
"errors"
"fmt"
)
// IntegrityError marks a finished download whose bytes don't match what was
// expected: a truncated / short file, a checksum/par2 failure, or an on-disk
// size below the advertised length. It is DISTINCT from a transport failure
// (network drop, dead tracker) — on an IntegrityError the manager re-downloads
// the SAME source a bounded number of times (a fresh, clean-start attempt), and
// only after exhausting the retries surfaces the task as damaged. This is the
// cross-backend safety net that guarantees "completed" never means a corrupt
// file (incident: 2026-06-15 debrid NFS write-back truncation — a 20 MB stub of
// a 394 MB file was marked completed because nothing re-checked the on-disk size
// after the page-cache write-back silently dropped ~374 MB).
type IntegrityError struct {
// Reason is a stable short code surfaced to the web / logs:
// "truncated", "size_mismatch", "empty", "par2_failed", "flush_failed".
Reason string
Message string // human-readable detail
}
func (e *IntegrityError) Error() string {
if e.Message != "" {
return e.Message
}
return "integrity check failed: " + e.Reason
}
// IsIntegrity reports whether err is (or wraps) an IntegrityError.
func IsIntegrity(err error) bool {
var ie *IntegrityError
return errors.As(err, &ie)
}
// integrityErr builds an IntegrityError with a printf-style message.
func integrityErr(reason, format string, args ...any) *IntegrityError {
return &IntegrityError{Reason: reason, Message: fmt.Sprintf(format, args...)}
}