Fix Issues #4, #5, #6: LXC-Kompatibilität, WorkingDirectory, GS-Backports

- #4: LXC/Container Drop-in (lxc-compat.conf) deaktiviert systemd-Hardening;
  Installer erkennt Container automatisch und bietet Drop-in an
- #5: WorkingDirectory=/opt/pdf-ocr-hotfolder in Template-Unit ergänzt
- #6: Installer bietet auf Debian 12 bei betroffenen GS-Versionen
  automatisch bookworm-backports Upgrade an (statt nur Warnung)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

.gitignore

@@ -4,6 +4,7 @@ __pycache__/
 venv/
 env/
 .venv/
+.pytest_cache/
 *.egg-info/
 build/
 dist/

CHANGELOG.md

@@ -1,5 +1,55 @@
 # Changelog
 
+## [0.3.1] - 2026-04-10
+
+### Fixed
+- **Issue #4**: LXC/Container-Kompatibilität — systemd-Hardening (`PrivateTmp`, `ProtectSystem`, etc.)
+  verursacht Error 226/NAMESPACE in LXC-Containern. Installer erkennt Container-Umgebung automatisch
+  und bietet ein Drop-in an. Zusätzlich liegt `systemd/lxc-compat.conf` als Vorlage im Repo.
+- **Issue #5**: `WorkingDirectory=/opt/pdf-ocr-hotfolder` in der systemd Template-Unit ergänzt —
+  ohne diesen Eintrag konnte das Python-Modul nicht gefunden werden.
+- **Issue #6**: Auf Debian 12 bietet der Installer bei betroffenen Ghostscript-Versionen (10.0.0–10.02.0)
+  jetzt automatisch an, bookworm-backports zu aktivieren und GS zu upgraden (statt nur zu warnen).
+
+## [0.3.0] - 2026-04-09
+
+### Added
+- Neue Config-Sektion `[output]` mit:
+  - `name_mode` — Platzierung des Tags im Dateinamen: `"prefix"`, `"suffix"` (vor Extension), `"none"`
+  - `name_tag` — verbatim einzufügender String, z.B. `"OCR_"` oder `"_OCR"`
+  - `original_on_success` — `"delete"` (alter Default) oder `"archive"`
+  - `archive_dir` — Zielverzeichnis für `"archive"`, mit Kollisions-Schutz (Timestamp-Suffix)
+- Runtime-Validierung der Output-Config in `check_output_config()`
+- 20 neue Tests für `build_output_name()`, `check_output_config()` und `process_pdf()`
+  mit allen Kombinationen aus Modus + Original-Behandlung
+
+### Changed
+- `process_pdf()` nimmt jetzt `output_cfg: OutputConfig` als Pflicht-Argument
+
+## [0.2.2] - 2026-04-09
+
+### Fixed
+- **Issue #3**: Ghostscript 10.0.0–10.02.0 (Debian 12 default) zerschießen OCR mit PDF/A + `skip_text=true`.
+  - `config.example.toml`: `pdfa_level = ""` als sicherer Default
+  - Runtime-Preflight: Prüft `gs --version` wenn `pdfa_level` gesetzt ist, bricht mit klarer Fehlermeldung ab
+  - `install.sh`: warnt bei betroffenen GS-Versionen mit Upgrade-Hinweis auf bookworm-backports
+
+### Added
+- `is_ghostscript_broken()` / `detect_ghostscript_version()` in `pdf_ocr_hotfolder.service`
+- 19 weitere pytest-Tests für GS-Versions-Detection (parametrisiert) und Preflight-Kombinationen
+
+## [0.2.1] - 2026-04-09
+
+### Fixed
+- **Issue #1**: Preflight-Check beim Start prüft jetzt `tesseract` und `gs` (Ghostscript). Fehlt eine Abhängigkeit, beendet sich der Service sofort mit Exit-Code 2 und klarer Fehlermeldung statt erst bei der ersten Datei.
+- **Issue #2**: `--once`-Modus liefert jetzt Exit-Code `1`, sobald **mindestens ein** PDF fehlgeschlagen ist. Exit-Code `0` nur bei vollständigem Erfolg (inkl. "keine Dateien vorhanden"). Exit-Code `2` bei Preflight-Fehler.
+
+### Added
+- Public API: `HotfolderService.run_once()`, `.success_count`, `.error_count`, `.ensure_dirs()`
+- `check_preflight()` / `PreflightError` in `pdf_ocr_hotfolder.service`
+- pytest-Test-Suite (`tests/`) mit 11 Tests — deckt alle Szenarien aus Issue #1 und #2 ab
+- `ocrmypdf`-Import in `processor.py` ist jetzt lazy (Tests ohne ocrmypdf-Installation möglich)
+
 ## [0.2.0] - 2026-04-08
 
 ### Added

README.md

@@ -89,6 +89,22 @@ max_workers = 2          # parallele PDFs
 timeout = 1800
 ```
 
+### `[output]`
+```toml
+# Dateiname im outgoing/:
+#   "prefix" → OCR_scan.pdf
+#   "suffix" → scan_OCR.pdf (vor der Extension)
+#   "none"   → scan.pdf (unverändert)
+name_mode = "prefix"
+name_tag  = "OCR_"
+
+# Nach erfolgreichem OCR mit dem Original:
+#   "delete"  → löschen
+#   "archive" → in archive_dir verschieben
+original_on_success = "delete"
+archive_dir = ""     # absoluter Pfad, Pflicht bei "archive"
+```
+
 ### `[upload.nextcloud]`
 ```toml
 enabled = true
@@ -174,6 +190,24 @@ Service-User braucht **rw** auf alle vier Verzeichnisse unter `/var/lib/pdf-ocr-
 sudo chown -R DOMAIN\\scanuser:DOMAIN\\scangroup /var/lib/pdf-ocr-hotfolder
 ```
 
+### LXC/Container: Error 226/NAMESPACE
+In LXC-Containern schlagen systemd-Hardening-Optionen fehl. Der Installer erkennt Container automatisch und bietet ein Drop-in an. Manuell:
+```bash
+sudo mkdir -p /etc/systemd/system/pdf-ocr-hotfolder@.service.d/
+sudo cp /opt/pdf-ocr-hotfolder/systemd/lxc-compat.conf \
+    /etc/systemd/system/pdf-ocr-hotfolder@.service.d/
+sudo systemctl daemon-reload
+sudo systemctl restart 'pdf-ocr-hotfolder@*'
+```
+
+### Ghostscript PDF/A-Bug auf Debian 12
+GS 10.00.0–10.02.0 (Debian 12 Default) zerstört OCR bei `pdfa_level` + `skip_text=true`. Der Installer bietet automatisch bookworm-backports an. Manuell:
+```bash
+echo 'deb http://deb.debian.org/debian bookworm-backports main' | \
+    sudo tee /etc/apt/sources.list.d/bookworm-backports.list
+sudo apt update && sudo apt install -t bookworm-backports ghostscript
+```
+
 ### veraPDF-Validierung schlägt immer fehl
 veraPDF binary prüfen (`[verapdf].binary`). Wenn nicht zwingend gebraucht: `enabled = false`.
 
@@ -205,5 +239,5 @@ MIT — © Sonith UG
 
 ---
 
-**Version:** 0.2.0
+**Version:** 0.3.1
 **Repo:** https://gitea.sonith.de/sonith_ug/pdf-ocr-hotfolder

VERSION

@@ -1 +1 @@
-0.2.0
+0.3.1

config.example.toml

@@ -21,7 +21,10 @@ skip_text = true
 # Auflösung für gerasterte Seiten
 oversample = 300
 # PDF/A-Konformitätsstufe ("1", "2", "3" oder leer für keinen PDF/A-Output)
-pdfa_level = "2"
+# ACHTUNG: Ghostscript 10.0.0 bis 10.02.0 (Debian 12 default!) haben einen Bug,
+# der mit pdfa_level + skip_text=true ocrmypdf komplett blockiert.
+# Sicherer Default ist "" — nur auf "1"/"2"/"3" setzen, wenn gs >= 10.02.1 installiert ist.
+pdfa_level = ""
 # Schiefe Scans automatisch begradigen
 deskew = true
 # Hintergrund säubern
@@ -31,6 +34,22 @@ max_workers = 2
 # Timeout pro PDF in Sekunden
 timeout = 1800
 
+[output]
+# Wie soll die Ziel-Datei im outgoing/-Ordner benannt werden?
+#   "prefix" : name_tag wird vor den Dateinamen gestellt  (OCR_scan.pdf)
+#   "suffix" : name_tag wird vor die Extension gestellt   (scan_OCR.pdf)
+#   "none"   : Dateiname bleibt wie das Original
+name_mode = "prefix"
+# Verbatim einzufügender String. Leerer String = kein Tag (wie mode="none").
+# Beispiele: "OCR_", "[OCR]_", "_OCR", "_searchable"
+name_tag = "OCR_"
+# Was passiert mit dem Original, wenn OCR erfolgreich war?
+#   "delete"  : Original wird gelöscht (alter Standard)
+#   "archive" : Original wird in archive_dir verschoben
+original_on_success = "delete"
+# Absoluter Pfad; nur relevant wenn original_on_success = "archive"
+archive_dir = ""
+
 [verapdf]
 # PDF/A-Validierung (optional)
 enabled = false

install.sh

@@ -52,6 +52,60 @@ install_base() {
         icc-profiles-free ca-certificates curl
     log_info "System-Pakete ok ✓"
 
+    # Ghostscript-Versions-Check (Issue #3 + Issue #6)
+    if command -v gs >/dev/null 2>&1; then
+        GS_VER="$(gs --version 2>/dev/null || echo 0.0)"
+        log_info "Ghostscript: $GS_VER"
+        case "$GS_VER" in
+            10.0.0|10.00.0|10.01.*|10.02.0)
+                echo
+                log_warn "═══════════════════════════════════════════════════════════════"
+                log_warn "Ghostscript $GS_VER ist vom PDF/A-Bug betroffen (10.0.0–10.02.0)."
+                log_warn "Mit pdfa_level + skip_text=true kann ocrmypdf KEINE PDFs verarbeiten."
+                log_warn "═══════════════════════════════════════════════════════════════"
+                echo
+                # Prüfe ob Debian bookworm (12) — Backports anbieten
+                if grep -q 'bookworm' /etc/os-release 2>/dev/null; then
+                    read -r -p "Ghostscript via bookworm-backports upgraden? [J/n]: " UPGRADE_GS
+                    UPGRADE_GS="${UPGRADE_GS:-J}"
+                    if [[ "$UPGRADE_GS" =~ ^[JjYy]$ ]]; then
+                        log_info "Aktiviere bookworm-backports..."
+                        if ! grep -q 'bookworm-backports' /etc/apt/sources.list /etc/apt/sources.list.d/*.list 2>/dev/null; then
+                            echo 'deb http://deb.debian.org/debian bookworm-backports main' \
+                                > /etc/apt/sources.list.d/bookworm-backports.list
+                            apt-get update -qq
+                        fi
+                        apt-get install -y -t bookworm-backports ghostscript
+                        GS_VER_NEW="$(gs --version 2>/dev/null || echo '?')"
+                        log_info "Ghostscript aktualisiert: $GS_VER → $GS_VER_NEW ✓"
+                    else
+                        log_warn "Workaround: In der Config [ocr].pdfa_level = \"\" setzen (Default ab v0.2.2)"
+                    fi
+                else
+                    log_warn "Kein Debian bookworm erkannt — manuelles Upgrade nötig."
+                    log_warn "Workaround: In der Config [ocr].pdfa_level = \"\" setzen (Default ab v0.2.2)"
+                fi
+                echo
+                ;;
+        esac
+    fi
+
+    # LXC/Container-Erkennung (Issue #4)
+    if systemd-detect-virt --container -q 2>/dev/null; then
+        VIRT_TYPE="$(systemd-detect-virt --container 2>/dev/null || echo 'container')"
+        log_warn "Container-Umgebung erkannt ($VIRT_TYPE)."
+        log_warn "systemd-Hardening kann in Containern fehlschlagen (Error 226/NAMESPACE)."
+        read -r -p "LXC-Kompatibilitäts-Drop-in installieren? [J/n]: " LXC_FIX
+        LXC_FIX="${LXC_FIX:-J}"
+        if [[ "$LXC_FIX" =~ ^[JjYy]$ ]]; then
+            local LXC_DROPIN_DIR="/etc/systemd/system/pdf-ocr-hotfolder@.service.d"
+            mkdir -p "$LXC_DROPIN_DIR"
+            cp "$REPO_DIR/systemd/lxc-compat.conf" "$LXC_DROPIN_DIR/lxc-compat.conf"
+            systemctl daemon-reload
+            log_info "LXC-Kompatibilitäts-Drop-in installiert ✓"
+        fi
+    fi
+
     log_step "Default-User '$DEFAULT_USER' prüfen"
     if id "$DEFAULT_USER" &>/dev/null; then
         log_info "'$DEFAULT_USER' existiert bereits"

pdf_ocr_hotfolder/__init__.py

@@ -1,3 +1,3 @@
 """PDF OCR Hotfolder — Scanner-PDFs automatisch durchsuchbar machen."""
 
-__version__ = "0.1.0"
+__version__ = "0.3.1"

pdf_ocr_hotfolder/__main__.py

@@ -8,7 +8,7 @@ from pathlib import Path
 
 from . import __version__
 from .config import load_config
-from .service import HotfolderService
+from .service import HotfolderService, PreflightError
 
 
 def _setup_logging(level: str) -> None:
@@ -40,14 +40,20 @@ def main() -> int:
     _setup_logging(cfg.log_level)
 
     service = HotfolderService(cfg)
+
     if args.once:
-        service._ensure_dirs()  # noqa: SLF001
+        try:
-        service._scan_existing()  # noqa: SLF001
+            errors = service.run_once()
-        service._executor.shutdown(wait=True)  # noqa: SLF001
+        except PreflightError as e:
-        return 0
+            print(f"FEHLER: {e}", file=sys.stderr)
+            return 2
+        return 1 if errors > 0 else 0
 
     try:
         service.run()
+    except PreflightError as e:
+        print(f"FEHLER: {e}", file=sys.stderr)
+        return 2
     except KeyboardInterrupt:
         pass
     return 0

pdf_ocr_hotfolder/config.py

@@ -28,6 +28,18 @@ class OcrConfig:
     timeout: int = 1800
 
 
+@dataclass
+class OutputConfig:
+    # "prefix" | "suffix" | "none"
+    name_mode: str = "prefix"
+    # Tag-String, verbatim eingefügt (Leerstring = kein Tag)
+    name_tag: str = "OCR_"
+    # "delete" | "archive"
+    original_on_success: str = "delete"
+    # Absoluter Pfad; Pflicht wenn original_on_success == "archive"
+    archive_dir: str = ""
+
+
 @dataclass
 class VeraPdfConfig:
     enabled: bool = False
@@ -79,6 +91,7 @@ class EmailNotify:
 class Config:
     paths: Paths
     ocr: OcrConfig
+    output: OutputConfig
     verapdf: VeraPdfConfig
     folder: FolderUpload
     nextcloud: NextcloudUpload
@@ -109,6 +122,8 @@ def load_config(path: str | Path) -> Config:
 
     ocr = OcrConfig(**{k: v for k, v in _section(data, "ocr").items()
                        if k in OcrConfig.__annotations__})
+    output = OutputConfig(**{k: v for k, v in _section(data, "output").items()
+                             if k in OutputConfig.__annotations__})
     verapdf = VeraPdfConfig(**{k: v for k, v in _section(data, "verapdf").items()
                                if k in VeraPdfConfig.__annotations__})
     folder = FolderUpload(**{k: v for k, v in _section(data, "upload", "folder").items()
@@ -123,7 +138,7 @@ def load_config(path: str | Path) -> Config:
     log_level = _section(data, "logging").get("level", "INFO")
 
     return Config(
-        paths=paths, ocr=ocr, verapdf=verapdf,
+        paths=paths, ocr=ocr, output=output, verapdf=verapdf,
         folder=folder, nextcloud=nextcloud, sftp=sftp, email=email,
         log_level=log_level,
     )

pdf_ocr_hotfolder/processor.py

@@ -7,13 +7,37 @@ import subprocess
 from dataclasses import dataclass
 from pathlib import Path
 
-import ocrmypdf
+from .config import OcrConfig, OutputConfig, VeraPdfConfig
-
-from .config import OcrConfig, VeraPdfConfig
 
 log = logging.getLogger(__name__)
 
 
+def build_output_name(src_name: str, mode: str, tag: str) -> str:
+    """Erzeugt den Ziel-Dateinamen für ein OCR-PDF.
+
+    Args:
+        src_name: Original-Dateiname (z.B. "scan.pdf")
+        mode: "prefix" | "suffix" | "none"
+        tag: Einzufügender String (verbatim, leer = kein Tag)
+
+    Beispiele:
+        prefix "OCR_":  "scan.pdf"  -> "OCR_scan.pdf"
+        suffix "_OCR":  "scan.pdf"  -> "scan_OCR.pdf"
+        suffix "_OCR":  "scan.tar.gz.pdf" -> "scan.tar.gz_OCR.pdf"
+        none:           "scan.pdf"  -> "scan.pdf"
+    """
+    if mode == "none" or not tag:
+        return src_name
+    if mode == "prefix":
+        return f"{tag}{src_name}"
+    if mode == "suffix":
+        # Nur die letzte Extension abspalten, sonst "foo.bar.pdf" kaputt gemacht
+        p = Path(src_name)
+        stem, ext = p.stem, p.suffix
+        return f"{stem}{tag}{ext}"
+    raise ValueError(f"Unbekannter name_mode: {mode!r}")
+
+
 @dataclass
 class ProcessResult:
     source: Path
@@ -25,6 +49,8 @@ class ProcessResult:
 
 def run_ocr(src: Path, dst: Path, cfg: OcrConfig) -> None:
     """Führt ocrmypdf als Library-Call aus (kein Subprozess-Overhead)."""
+    import ocrmypdf  # lazy, damit Tests ohne ocrmypdf laufen
+
     kwargs: dict = {
         "language": cfg.languages,
         "jobs": cfg.jobs,
@@ -71,11 +97,13 @@ def process_pdf(
     error_dir: Path,
     ocr_cfg: OcrConfig,
     vera_cfg: VeraPdfConfig,
+    output_cfg: OutputConfig,
 ) -> ProcessResult:
     """Verarbeitet eine einzelne PDF: move→OCR→validate→outgoing/error."""
+    out_name = build_output_name(src.name, output_cfg.name_mode, output_cfg.name_tag)
     work_src = working_dir / src.name
-    work_out = working_dir / f"OCR_{src.name}"
+    work_out = working_dir / f"__ocr_{out_name}"  # Temp-Name, damit er != src.name ist
-    final_out = outgoing_dir / f"OCR_{src.name}"
+    final_out = outgoing_dir / out_name
 
     try:
         shutil.move(str(src), str(work_src))
@@ -100,10 +128,38 @@ def process_pdf(
 
     outgoing_dir.mkdir(parents=True, exist_ok=True)
     shutil.move(str(work_out), str(final_out))
-    work_src.unlink(missing_ok=True)
+    _dispose_original(work_src, src.name, output_cfg)
     return ProcessResult(src, final_out, True, verapdf_passed=vera_ok)
 
 
+def _dispose_original(work_src: Path, original_name: str, cfg: OutputConfig) -> None:
+    """Entsorgt das Original nach erfolgreichem OCR — löschen oder archivieren."""
+    if not work_src.exists():
+        return
+    mode = cfg.original_on_success
+    if mode == "delete":
+        work_src.unlink(missing_ok=True)
+        return
+    if mode == "archive":
+        if not cfg.archive_dir:
+            log.error("original_on_success=archive aber archive_dir ist leer — lösche stattdessen")
+            work_src.unlink(missing_ok=True)
+            return
+        archive = Path(cfg.archive_dir)
+        archive.mkdir(parents=True, exist_ok=True)
+        dest = archive / original_name
+        # Bei Namens-Kollision mit Timestamp umbenennen
+        if dest.exists():
+            from datetime import datetime
+            ts = datetime.now().strftime("%Y%m%d-%H%M%S")
+            dest = archive / f"{dest.stem}_{ts}{dest.suffix}"
+        shutil.move(str(work_src), str(dest))
+        log.info("Original archiviert: %s", dest)
+        return
+    log.warning("Unbekannter original_on_success=%r — lösche stattdessen", mode)
+    work_src.unlink(missing_ok=True)
+
+
 def _move_to_error(p: Path, error_dir: Path) -> None:
     error_dir.mkdir(parents=True, exist_ok=True)
     try:

pdf_ocr_hotfolder/service.py

@@ -2,7 +2,10 @@
 from __future__ import annotations
 
 import logging
+import re
+import shutil
 import signal
+import subprocess
 import threading
 import time
 from concurrent.futures import Future, ThreadPoolExecutor
@@ -18,6 +21,98 @@ from .uploaders import notify_email, upload_folder, upload_nextcloud, upload_sft
 log = logging.getLogger(__name__)
 
 
+class PreflightError(RuntimeError):
+    """Erforderliche externe Binaries fehlen."""
+
+
+# Pflicht-Binaries für ocrmypdf
+_REQUIRED_BINARIES = ("tesseract", "gs")
+
+# Ghostscript-Versionen mit bekanntem PDF/A+skip_text Bug (Issue #3):
+# 10.0.0 .. 10.02.0 (inklusive). Ab 10.02.1 wieder nutzbar.
+_GS_BROKEN_MIN = (10, 0, 0)
+_GS_BROKEN_MAX = (10, 2, 0)
+
+
+def _parse_version(text: str) -> tuple[int, ...] | None:
+    """Extrahiert die erste X.Y[.Z] Version aus einem String."""
+    m = re.search(r"(\d+)\.(\d+)(?:\.(\d+))?", text)
+    if not m:
+        return None
+    return tuple(int(x) if x is not None else 0 for x in m.groups())
+
+
+def is_ghostscript_broken(version: str | None) -> bool:
+    """Prüft, ob eine Ghostscript-Version vom PDF/A+skip_text Bug betroffen ist.
+
+    Betrifft 10.0.0 bis einschließlich 10.02.0. Ab 10.02.1 wieder sicher.
+    """
+    if not version:
+        return False
+    parsed = _parse_version(version)
+    if parsed is None:
+        return False
+    # Auf 3-Tupel normalisieren
+    while len(parsed) < 3:
+        parsed = parsed + (0,)
+    parsed = parsed[:3]
+    return _GS_BROKEN_MIN <= parsed <= _GS_BROKEN_MAX
+
+
+def detect_ghostscript_version() -> str | None:
+    """Ruft `gs --version` auf und gibt den Versionsstring zurück (oder None)."""
+    gs = shutil.which("gs")
+    if gs is None:
+        return None
+    try:
+        result = subprocess.run([gs, "--version"], capture_output=True,
+                                text=True, timeout=5)
+    except (OSError, subprocess.TimeoutExpired):
+        return None
+    return result.stdout.strip() or None
+
+
+def check_output_config(mode: str, archive_dir: str) -> None:
+    """Validiert die [output]-Section. Wirft PreflightError bei Problemen."""
+    valid_modes = {"delete", "archive"}
+    if mode not in valid_modes:
+        raise PreflightError(
+            f"[output].original_on_success={mode!r} ungültig. "
+            f"Erlaubt: {sorted(valid_modes)}"
+        )
+    if mode == "archive" and not archive_dir:
+        raise PreflightError(
+            "[output].original_on_success='archive' erfordert [output].archive_dir"
+        )
+
+
+def check_preflight(pdfa_level: str = "") -> None:
+    """Prüft externe Abhängigkeiten.
+
+    - Tesseract und Ghostscript müssen im PATH sein
+    - Bei gesetztem pdfa_level wird die Ghostscript-Version gegen den
+      bekannten 10.0.0–10.02.0 Bug geprüft
+
+    Wirft PreflightError bei fehlenden Binaries oder unsicherem Ghostscript.
+    """
+    missing = [b for b in _REQUIRED_BINARIES if shutil.which(b) is None]
+    if missing:
+        raise PreflightError(
+            "Fehlende Abhängigkeiten: " + ", ".join(missing)
+            + ". Bitte installieren: sudo apt install tesseract-ocr ghostscript"
+        )
+
+    if pdfa_level:
+        gs_version = detect_ghostscript_version()
+        if is_ghostscript_broken(gs_version):
+            raise PreflightError(
+                f"Ghostscript {gs_version} ist mit pdfa_level='{pdfa_level}' nicht "
+                "kompatibel (bekannter Bug in 10.0.0–10.02.0). "
+                "Entweder ghostscript auf >=10.02.1 upgraden (z.B. via bookworm-backports) "
+                "oder in der Config [ocr].pdfa_level = \"\" setzen."
+            )
+
+
 def _is_pdf(path: Path) -> bool:
     return path.suffix.lower() == ".pdf" and path.is_file()
 
@@ -70,10 +165,20 @@ class HotfolderService:
         self._stop = threading.Event()
         self._inflight: set[str] = set()
         self._lock = threading.Lock()
+        self._success_count = 0
+        self._error_count = 0
+
+    @property
+    def success_count(self) -> int:
+        return self._success_count
+
+    @property
+    def error_count(self) -> int:
+        return self._error_count
 
     # ---- Setup ----
 
-    def _ensure_dirs(self) -> None:
+    def ensure_dirs(self) -> None:
         for p in (self.cfg.paths.incoming, self.cfg.paths.outgoing,
                   self.cfg.paths.working, self.cfg.paths.error):
             p.mkdir(parents=True, exist_ok=True)
@@ -81,7 +186,10 @@ class HotfolderService:
     # ---- Lifecycle ----
 
     def run(self) -> None:
-        self._ensure_dirs()
+        check_preflight(self.cfg.ocr.pdfa_level)
+        check_output_config(self.cfg.output.original_on_success,
+                            self.cfg.output.archive_dir)
+        self.ensure_dirs()
         self._scan_existing()
 
         self._observer = Observer()
@@ -98,6 +206,22 @@ class HotfolderService:
         finally:
             self.shutdown()
 
+    def run_once(self) -> int:
+        """Verarbeitet alle bereits im incoming-Ordner liegenden PDFs und beendet sich.
+
+        Returns:
+            Anzahl fehlgeschlagener PDFs (0 = alles ok).
+        """
+        check_preflight(self.cfg.ocr.pdfa_level)
+        check_output_config(self.cfg.output.original_on_success,
+                            self.cfg.output.archive_dir)
+        self.ensure_dirs()
+        self._scan_existing()
+        self._executor.shutdown(wait=True)
+        log.info("One-shot fertig: %d ok, %d Fehler",
+                 self._success_count, self._error_count)
+        return self._error_count
+
     def shutdown(self) -> None:
         log.info("Shutdown läuft...")
         if self._observer:
@@ -148,8 +272,15 @@ class HotfolderService:
             error_dir=self.cfg.paths.error,
             ocr_cfg=self.cfg.ocr,
             vera_cfg=self.cfg.verapdf,
+            output_cfg=self.cfg.output,
         )
 
+        with self._lock:
+            if result.success:
+                self._success_count += 1
+            else:
+                self._error_count += 1
+
         if result.success:
             self._dispatch_uploads(result.output)
         self._notify(result)

systemd/lxc-compat.conf

@@ -0,0 +1,10 @@
+# Drop-in für LXC/Container-Betrieb
+# Kopieren nach: /etc/systemd/system/pdf-ocr-hotfolder@.service.d/lxc-compat.conf
+# Danach: systemctl daemon-reload && systemctl restart 'pdf-ocr-hotfolder@*'
+
+[Service]
+PrivateTmp=false
+ProtectSystem=false
+ProtectKernelTunables=false
+ProtectKernelModules=false
+ProtectControlGroups=false

systemd/pdf-ocr-hotfolder@.service

@@ -7,6 +7,7 @@ Wants=network-online.target
 Type=simple
 User=pdfocr
 Group=pdfocr
+WorkingDirectory=/opt/pdf-ocr-hotfolder
 ExecStart=/opt/pdf-ocr-hotfolder/venv/bin/python -m pdf_ocr_hotfolder --config /etc/pdf-ocr-hotfolder/%i.toml
 Restart=on-failure
 RestartSec=5

tests/conftest.py

@@ -0,0 +1,54 @@
+"""Gemeinsame pytest-Fixtures."""
+from __future__ import annotations
+
+from pathlib import Path
+
+import pytest
+
+from pdf_ocr_hotfolder.config import (
+    Config,
+    EmailNotify,
+    FolderUpload,
+    NextcloudUpload,
+    OcrConfig,
+    OutputConfig,
+    Paths,
+    SftpUpload,
+    VeraPdfConfig,
+)
+
+
+@pytest.fixture
+def tmp_config(tmp_path: Path) -> Config:
+    """Minimal-Config mit tmp_path-Verzeichnissen, alle Uploads deaktiviert."""
+    paths = Paths(
+        incoming=tmp_path / "incoming",
+        outgoing=tmp_path / "outgoing",
+        working=tmp_path / "working",
+        error=tmp_path / "error",
+    )
+    for p in (paths.incoming, paths.outgoing, paths.working, paths.error):
+        p.mkdir(parents=True, exist_ok=True)
+
+    return Config(
+        paths=paths,
+        ocr=OcrConfig(max_workers=1),
+        output=OutputConfig(),
+        verapdf=VeraPdfConfig(enabled=False),
+        folder=FolderUpload(enabled=False),
+        nextcloud=NextcloudUpload(enabled=False),
+        sftp=SftpUpload(enabled=False),
+        email=EmailNotify(enabled=False),
+        log_level="DEBUG",
+    )
+
+
+@pytest.fixture
+def dummy_pdf(tmp_config: Config) -> Path:
+    """Legt eine Datei mit .pdf-Extension im incoming-Ordner ab.
+
+    Achtung: kein echtes PDF. Für Tests wird `process_pdf` gemockt.
+    """
+    pdf = tmp_config.paths.incoming / "test.pdf"
+    pdf.write_bytes(b"%PDF-1.4 fake\n")
+    return pdf

tests/test_ghostscript_version.py

@@ -0,0 +1,72 @@
+"""Tests für Issue #3: Ghostscript 10.0.0–10.02.0 PDF/A-Bug-Erkennung."""
+from __future__ import annotations
+
+from unittest.mock import patch
+
+import pytest
+
+from pdf_ocr_hotfolder.service import (
+    PreflightError,
+    check_preflight,
+    is_ghostscript_broken,
+)
+
+
+@pytest.mark.parametrize("version,expected", [
+    # Betroffene Versionen
+    ("10.0.0", True),
+    ("10.00.0", True),
+    ("10.01.0", True),
+    ("10.01.1", True),
+    ("10.01.2", True),
+    ("10.02.0", True),
+    # Sichere Versionen
+    ("10.02.1", False),
+    ("10.03.0", False),
+    ("10.04.0", False),
+    ("11.0.0", False),
+    ("9.56.1", False),  # Debian 11 / Ubuntu 22.04
+    ("9.55.0", False),
+    # Edge cases
+    ("", False),
+    (None, False),
+    ("garbage", False),
+])
+def test_is_ghostscript_broken(version, expected) -> None:
+    assert is_ghostscript_broken(version) is expected
+
+
+def test_check_preflight_without_pdfa_passes_with_broken_gs() -> None:
+    """Ohne pdfa_level darf der betroffene GS verwendet werden."""
+    with patch("pdf_ocr_hotfolder.service.shutil.which", return_value="/usr/bin/fake"), \
+         patch("pdf_ocr_hotfolder.service.detect_ghostscript_version",
+               return_value="10.0.0"):
+        check_preflight(pdfa_level="")  # darf nicht werfen
+
+
+def test_check_preflight_with_pdfa_fails_on_broken_gs() -> None:
+    """Mit pdfa_level + kaputtem GS → PreflightError mit hilfreicher Meldung."""
+    with patch("pdf_ocr_hotfolder.service.shutil.which", return_value="/usr/bin/fake"), \
+         patch("pdf_ocr_hotfolder.service.detect_ghostscript_version",
+               return_value="10.0.0"):
+        with pytest.raises(PreflightError, match="Ghostscript 10.0.0"):
+            check_preflight(pdfa_level="2")
+
+
+def test_check_preflight_with_pdfa_passes_on_fixed_gs() -> None:
+    """Mit pdfa_level + gefixtem GS → ok."""
+    with patch("pdf_ocr_hotfolder.service.shutil.which", return_value="/usr/bin/fake"), \
+         patch("pdf_ocr_hotfolder.service.detect_ghostscript_version",
+               return_value="10.02.1"):
+        check_preflight(pdfa_level="2")  # darf nicht werfen
+
+
+def test_default_config_pdfa_level_is_empty() -> None:
+    """Default-Config der Beispiel-Datei soll pdfa_level='' enthalten (Issue #3)."""
+    from pathlib import Path
+    import tomllib
+    cfg_path = Path(__file__).parent.parent / "config.example.toml"
+    with cfg_path.open("rb") as f:
+        data = tomllib.load(f)
+    assert data["ocr"]["pdfa_level"] == "", \
+        "config.example.toml muss pdfa_level='' als sicheren Default haben"

tests/test_once_exit_code.py

@@ -0,0 +1,96 @@
+"""Tests für Issue #2: --once Modus muss Exit-Code != 0 bei Fehlern liefern."""
+from __future__ import annotations
+
+from pathlib import Path
+from unittest.mock import patch
+
+from pdf_ocr_hotfolder.processor import ProcessResult
+from pdf_ocr_hotfolder.service import HotfolderService
+
+
+def _fake_success(src: Path, working_dir, outgoing_dir, error_dir, **kwargs):
+    out = outgoing_dir / f"OCR_{src.name}"
+    out.parent.mkdir(parents=True, exist_ok=True)
+    out.write_bytes(b"%PDF-1.4 ocr\n")
+    src.unlink(missing_ok=True)
+    return ProcessResult(src, out, True)
+
+
+def _fake_failure(src: Path, working_dir, outgoing_dir, error_dir, **kwargs):
+    error_dir.mkdir(parents=True, exist_ok=True)
+    dest = error_dir / src.name
+    src.rename(dest)
+    return ProcessResult(src, outgoing_dir / f"OCR_{src.name}", False,
+                         error="fake ocr failure")
+
+
+def _run(tmp_config, fake_process):
+    """Helper: führt run_once() mit gemocktem process_pdf und preflight aus."""
+    with patch("pdf_ocr_hotfolder.service.check_preflight", return_value=None), \
+         patch("pdf_ocr_hotfolder.service.process_pdf", side_effect=fake_process), \
+         patch("pdf_ocr_hotfolder.service._wait_until_stable", return_value=True):
+        service = HotfolderService(tmp_config)
+        try:
+            return service.run_once()
+        finally:
+            service._executor.shutdown(wait=False)
+
+
+def test_once_exit_0_when_no_files(tmp_config) -> None:
+    """Szenario: Keine PDFs vorhanden → Exit 0."""
+    errors = _run(tmp_config, _fake_success)
+    assert errors == 0
+
+
+def test_once_exit_0_when_all_success(tmp_config) -> None:
+    """Szenario: Alle PDFs erfolgreich → Exit 0."""
+    (tmp_config.paths.incoming / "a.pdf").write_bytes(b"%PDF-1.4\n")
+    (tmp_config.paths.incoming / "b.pdf").write_bytes(b"%PDF-1.4\n")
+
+    errors = _run(tmp_config, _fake_success)
+    assert errors == 0
+
+
+def test_once_exit_nonzero_when_all_fail(tmp_config) -> None:
+    """Szenario: Alle PDFs fehlgeschlagen → Exit != 0 (Issue #2)."""
+    (tmp_config.paths.incoming / "a.pdf").write_bytes(b"%PDF-1.4\n")
+    (tmp_config.paths.incoming / "b.pdf").write_bytes(b"%PDF-1.4\n")
+
+    errors = _run(tmp_config, _fake_failure)
+    assert errors == 2
+
+
+def test_once_exit_nonzero_when_some_fail(tmp_config) -> None:
+    """Szenario: Teilweise fehlgeschlagen → Exit != 0."""
+    (tmp_config.paths.incoming / "ok.pdf").write_bytes(b"%PDF-1.4\n")
+    (tmp_config.paths.incoming / "bad.pdf").write_bytes(b"%PDF-1.4\n")
+
+    def mixed(src, *args, **kwargs):
+        if "bad" in src.name:
+            return _fake_failure(src, *args, **kwargs)
+        return _fake_success(src, *args, **kwargs)
+
+    errors = _run(tmp_config, mixed)
+    assert errors == 1
+
+
+def test_counters_track_success_and_failure(tmp_config) -> None:
+    """success_count und error_count sollen korrekt mitzählen."""
+    (tmp_config.paths.incoming / "ok.pdf").write_bytes(b"%PDF-1.4\n")
+    (tmp_config.paths.incoming / "bad.pdf").write_bytes(b"%PDF-1.4\n")
+
+    def mixed(src, *args, **kwargs):
+        if "bad" in src.name:
+            return _fake_failure(src, *args, **kwargs)
+        return _fake_success(src, *args, **kwargs)
+
+    with patch("pdf_ocr_hotfolder.service.check_preflight", return_value=None), \
+         patch("pdf_ocr_hotfolder.service.process_pdf", side_effect=mixed), \
+         patch("pdf_ocr_hotfolder.service._wait_until_stable", return_value=True):
+        service = HotfolderService(tmp_config)
+        try:
+            service.run_once()
+            assert service.success_count == 1
+            assert service.error_count == 1
+        finally:
+            service._executor.shutdown(wait=False)

tests/test_output_naming.py

@@ -0,0 +1,190 @@
+"""Tests für Feature: konfigurierbare Dateinamen und Original-Behandlung."""
+from __future__ import annotations
+
+from pathlib import Path
+from unittest.mock import patch
+
+import pytest
+
+from pdf_ocr_hotfolder.config import OcrConfig, OutputConfig, VeraPdfConfig
+from pdf_ocr_hotfolder.processor import build_output_name, process_pdf
+from pdf_ocr_hotfolder.service import PreflightError, check_output_config
+
+
+# ---------------- build_output_name ----------------
+
+@pytest.mark.parametrize("src,mode,tag,expected", [
+    # prefix
+    ("scan.pdf",          "prefix", "OCR_",  "OCR_scan.pdf"),
+    ("scan.pdf",          "prefix", "[OCR] ", "[OCR] scan.pdf"),
+    # suffix (Tag vor Extension)
+    ("scan.pdf",          "suffix", "_OCR",  "scan_OCR.pdf"),
+    ("scan.pdf",          "suffix", "-ocr",  "scan-ocr.pdf"),
+    # none
+    ("scan.pdf",          "none",   "OCR_",  "scan.pdf"),
+    # leerer Tag = none
+    ("scan.pdf",          "prefix", "",      "scan.pdf"),
+    ("scan.pdf",          "suffix", "",      "scan.pdf"),
+    # Mehrfach-Punkte im Namen: nur letzte Extension zählt
+    ("rechnung.2026.pdf", "suffix", "_OCR",  "rechnung.2026_OCR.pdf"),
+    ("rechnung.2026.pdf", "prefix", "OCR_",  "OCR_rechnung.2026.pdf"),
+    # Name ohne Extension
+    ("NO_EXT",            "suffix", "_OCR",  "NO_EXT_OCR"),
+])
+def test_build_output_name(src, mode, tag, expected) -> None:
+    assert build_output_name(src, mode, tag) == expected
+
+
+def test_build_output_name_invalid_mode() -> None:
+    with pytest.raises(ValueError, match="name_mode"):
+        build_output_name("x.pdf", "bogus", "OCR_")
+
+
+# ---------------- check_output_config ----------------
+
+def test_check_output_config_delete_ok() -> None:
+    check_output_config("delete", "")  # ok
+
+
+def test_check_output_config_archive_requires_dir() -> None:
+    with pytest.raises(PreflightError, match="archive_dir"):
+        check_output_config("archive", "")
+
+
+def test_check_output_config_archive_with_dir_ok() -> None:
+    check_output_config("archive", "/var/archive")  # ok
+
+
+def test_check_output_config_invalid_mode() -> None:
+    with pytest.raises(PreflightError, match="ungültig"):
+        check_output_config("trash", "")
+
+
+# ---------------- process_pdf mit Original-Behandlung ----------------
+
+def _fake_ocr(src: Path, dst: Path, cfg: OcrConfig) -> None:
+    """Simuliert ocrmypdf: kopiert Inhalt, erzeugt Zieldatei."""
+    dst.write_bytes(b"%PDF-1.4 OCRed\n" + src.read_bytes())
+
+
+def _prepare(tmp_path: Path) -> dict:
+    dirs = {
+        "working": tmp_path / "working",
+        "outgoing": tmp_path / "outgoing",
+        "error": tmp_path / "error",
+        "archive": tmp_path / "archive",
+        "incoming": tmp_path / "incoming",
+    }
+    for d in dirs.values():
+        d.mkdir(parents=True, exist_ok=True)
+    src = dirs["incoming"] / "scan.pdf"
+    src.write_bytes(b"%PDF-1.4 original\n")
+    return {"src": src, **dirs}
+
+
+def test_process_pdf_prefix_delete(tmp_path: Path) -> None:
+    env = _prepare(tmp_path)
+    out_cfg = OutputConfig(name_mode="prefix", name_tag="OCR_",
+                           original_on_success="delete")
+    with patch("pdf_ocr_hotfolder.processor.run_ocr", side_effect=_fake_ocr):
+        result = process_pdf(
+            src=env["src"],
+            working_dir=env["working"],
+            outgoing_dir=env["outgoing"],
+            error_dir=env["error"],
+            ocr_cfg=OcrConfig(),
+            vera_cfg=VeraPdfConfig(enabled=False),
+            output_cfg=out_cfg,
+        )
+    assert result.success
+    assert (env["outgoing"] / "OCR_scan.pdf").exists()
+    # Original ist weg, weder in incoming noch in working
+    assert not env["src"].exists()
+    assert not (env["working"] / "scan.pdf").exists()
+
+
+def test_process_pdf_suffix_delete(tmp_path: Path) -> None:
+    env = _prepare(tmp_path)
+    out_cfg = OutputConfig(name_mode="suffix", name_tag="_OCR",
+                           original_on_success="delete")
+    with patch("pdf_ocr_hotfolder.processor.run_ocr", side_effect=_fake_ocr):
+        result = process_pdf(
+            src=env["src"],
+            working_dir=env["working"],
+            outgoing_dir=env["outgoing"],
+            error_dir=env["error"],
+            ocr_cfg=OcrConfig(),
+            vera_cfg=VeraPdfConfig(enabled=False),
+            output_cfg=out_cfg,
+        )
+    assert result.success
+    assert (env["outgoing"] / "scan_OCR.pdf").exists()
+
+
+def test_process_pdf_none_mode(tmp_path: Path) -> None:
+    env = _prepare(tmp_path)
+    out_cfg = OutputConfig(name_mode="none", name_tag="OCR_",
+                           original_on_success="delete")
+    with patch("pdf_ocr_hotfolder.processor.run_ocr", side_effect=_fake_ocr):
+        result = process_pdf(
+            src=env["src"],
+            working_dir=env["working"],
+            outgoing_dir=env["outgoing"],
+            error_dir=env["error"],
+            ocr_cfg=OcrConfig(),
+            vera_cfg=VeraPdfConfig(enabled=False),
+            output_cfg=out_cfg,
+        )
+    assert result.success
+    # Ausgang hat GLEICHEN Namen wie Original
+    assert (env["outgoing"] / "scan.pdf").exists()
+
+
+def test_process_pdf_archive_original(tmp_path: Path) -> None:
+    env = _prepare(tmp_path)
+    out_cfg = OutputConfig(name_mode="prefix", name_tag="OCR_",
+                           original_on_success="archive",
+                           archive_dir=str(env["archive"]))
+    with patch("pdf_ocr_hotfolder.processor.run_ocr", side_effect=_fake_ocr):
+        result = process_pdf(
+            src=env["src"],
+            working_dir=env["working"],
+            outgoing_dir=env["outgoing"],
+            error_dir=env["error"],
+            ocr_cfg=OcrConfig(),
+            vera_cfg=VeraPdfConfig(enabled=False),
+            output_cfg=out_cfg,
+        )
+    assert result.success
+    assert (env["outgoing"] / "OCR_scan.pdf").exists()
+    # Original liegt jetzt im Archiv
+    archived = env["archive"] / "scan.pdf"
+    assert archived.exists()
+    assert archived.read_bytes() == b"%PDF-1.4 original\n"
+
+
+def test_process_pdf_archive_name_collision(tmp_path: Path) -> None:
+    """Bei Namens-Kollision im Archiv wird Timestamp angehängt."""
+    env = _prepare(tmp_path)
+    # Vorhandene Kollisions-Datei
+    (env["archive"] / "scan.pdf").write_bytes(b"old")
+
+    out_cfg = OutputConfig(name_mode="prefix", name_tag="OCR_",
+                           original_on_success="archive",
+                           archive_dir=str(env["archive"]))
+    with patch("pdf_ocr_hotfolder.processor.run_ocr", side_effect=_fake_ocr):
+        process_pdf(
+            src=env["src"],
+            working_dir=env["working"],
+            outgoing_dir=env["outgoing"],
+            error_dir=env["error"],
+            ocr_cfg=OcrConfig(),
+            vera_cfg=VeraPdfConfig(enabled=False),
+            output_cfg=out_cfg,
+        )
+    # Alte Datei unverändert
+    assert (env["archive"] / "scan.pdf").read_bytes() == b"old"
+    # Neue Datei mit Timestamp-Suffix
+    archived = list(env["archive"].glob("scan_*.pdf"))
+    assert len(archived) == 1
+    assert archived[0].read_bytes() == b"%PDF-1.4 original\n"

tests/test_preflight.py

@@ -0,0 +1,75 @@
+"""Tests für Issue #1: Preflight-Check bei fehlendem Tesseract."""
+from __future__ import annotations
+
+import sys
+from unittest.mock import patch
+
+import pytest
+
+from pdf_ocr_hotfolder.service import (
+    HotfolderService,
+    PreflightError,
+    check_preflight,
+)
+
+
+def test_preflight_passes_when_all_binaries_present() -> None:
+    """Wenn tesseract + gs im PATH sind, darf kein Fehler fliegen."""
+    with patch("pdf_ocr_hotfolder.service.shutil.which", return_value="/usr/bin/fake"):
+        check_preflight()  # darf nicht werfen
+
+
+def test_preflight_fails_when_tesseract_missing() -> None:
+    """Fehlendes tesseract → PreflightError mit passender Meldung."""
+    def fake_which(name: str) -> str | None:
+        return None if name == "tesseract" else "/usr/bin/fake"
+
+    with patch("pdf_ocr_hotfolder.service.shutil.which", side_effect=fake_which):
+        with pytest.raises(PreflightError, match="tesseract"):
+            check_preflight()
+
+
+def test_preflight_fails_when_ghostscript_missing() -> None:
+    def fake_which(name: str) -> str | None:
+        return None if name == "gs" else "/usr/bin/fake"
+
+    with patch("pdf_ocr_hotfolder.service.shutil.which", side_effect=fake_which):
+        with pytest.raises(PreflightError, match="gs"):
+            check_preflight()
+
+
+def test_preflight_lists_all_missing_binaries() -> None:
+    """Bei mehreren fehlenden Binaries werden alle genannt."""
+    with patch("pdf_ocr_hotfolder.service.shutil.which", return_value=None):
+        with pytest.raises(PreflightError) as exc_info:
+            check_preflight()
+        msg = str(exc_info.value)
+        assert "tesseract" in msg
+        assert "gs" in msg
+
+
+def test_run_once_raises_preflight_error(tmp_config) -> None:
+    """HotfolderService.run_once() wirft PreflightError, wenn tesseract fehlt."""
+    service = HotfolderService(tmp_config)
+    try:
+        with patch("pdf_ocr_hotfolder.service.shutil.which", return_value=None):
+            with pytest.raises(PreflightError):
+                service.run_once()
+    finally:
+        service._executor.shutdown(wait=False)
+
+
+def test_main_returns_2_on_preflight_error(tmp_config, tmp_path, monkeypatch) -> None:
+    """CLI liefert Exit-Code 2 bei Preflight-Fehler (Issue #1 Szenario)."""
+    cfg_file = tmp_path / "cfg.toml"
+    cfg_file.write_text(f"""
+[paths]
+incoming = "{tmp_config.paths.incoming}"
+outgoing = "{tmp_config.paths.outgoing}"
+working  = "{tmp_config.paths.working}"
+error    = "{tmp_config.paths.error}"
+""")
+    monkeypatch.setattr(sys, "argv", ["pdf-ocr-hotfolder", "--config", str(cfg_file), "--once"])
+    with patch("pdf_ocr_hotfolder.service.shutil.which", return_value=None):
+        from pdf_ocr_hotfolder.__main__ import main
+        assert main() == 2