Bump version: v17.7.1

Since v16.4.3, OCRmyPDF forced pdfminer's read buffer to 256 MiB to work
around a pdfminer bug that mishandled tokens split across the buffer
boundary (gh #1361). On Windows this caused a severe performance
regression (gh #1662): CPython's BufferedReader.read(n) eagerly allocates
an n-byte buffer on every read, so pdfminer's thousands of seek+read
cycles each paid a ~30 ms 256 MiB allocation (this allocation is lazy and
effectively free on Linux). For a typical PDF the "Scanning contents"
phase went from ~5s on Linux to ~60s on Windows.

The underlying pdfminer bug was fixed upstream in pdfminer.six 20250327
(pdfminer/pdfminer.six#1030), with a follow-up for tokens split across
streams in 20260107 (pdfminer/pdfminer.six#1158). Remove the monkeypatch
entirely and raise the minimum pdfminer.six to 20260107 so we rely on the
upstream fix instead.

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>

bin/bump_version.py

@@ -18,8 +18,9 @@ import cyclopts
 from packaging.version import InvalidVersion, Version
 
 try:
-    from github import Github, GithubException
+    from github import Auth, Github, GithubException
 except ImportError:
+    Auth = None  # type: ignore
     Github = None  # type: ignore
     GithubException = Exception  # type: ignore
 
@@ -68,7 +69,7 @@ def validate_release_notes(new_version: str) -> bool:
 
 def get_github_client():
     """Get an authenticated GitHub client."""
-    if Github is None:
+    if Github is None or Auth is None:
         print(f"{RED}error:{OFF} PyGithub is not installed")
         print("       Install with: pip install PyGithub")
         return None
@@ -92,7 +93,7 @@ def get_github_client():
             return None
 
     try:
-        return Github(token)
+        return Github(auth=Auth.Token(token))
     except GithubException as e:
         print(f"{RED}error:{OFF} Failed to authenticate with GitHub: {e}")
         return None

docs/releasenotes/version17.md

@@ -3,6 +3,36 @@
 
 # v17
 
+## v17.7.1
+
+- Fixed a severe, Windows-specific performance regression in the "Scanning
+  contents" phase, most visible with `--redo-ocr` ({issue}`1662`). Since
+  v16.4.3, OCRmyPDF forced pdfminer's read buffer to 256 MiB to work around a
+  pdfminer bug that mishandled tokens split across the buffer boundary
+  ({issue}`1361`). On Windows, CPython's `BufferedReader.read()` eagerly
+  allocates a buffer of the requested size on every read, so the oversized
+  buffer made each of pdfminer's thousands of reads cost tens of milliseconds
+  (this allocation is lazy, and effectively free, on Linux). The underlying
+  pdfminer bug was fixed upstream in pdfminer.six 20250327
+  ([#1030](https://github.com/pdfminer/pdfminer.six/pull/1030)), with a
+  follow-up for tokens split across streams in 20260107
+  ([#1158](https://github.com/pdfminer/pdfminer.six/pull/1158)), so the
+  workaround has been removed and the minimum pdfminer.six version raised to
+  20260107.
+- The font discovery used to build the OCR text layer now finds variable fonts
+  such as `NotoSansArabic[wdth,wght].ttf`, the form shipped by Homebrew casks
+  and current Google Fonts releases. Previously only static `-Regular.ttf`/`.otf`
+  files were matched, so users who had installed the correct Noto font still got
+  the glyphless fallback and a "No font found" warning ({issue}`1652`).
+- Font discovery is now language-aware for CJK: each Chinese, Japanese, and
+  Korean language maps to its own per-language Noto family (NotoSansSC, TC, HK,
+  JP, KR), with the pan-CJK super font kept as a shared fallback, since the
+  per-language fonts are region subsets that may lack glyphs from other scripts.
+- The warning shown when no installed font has glyphs for some text was reworded
+  to explain the consequence — the text is still added as a searchable, copyable
+  layer but appears blank when highlighted in a viewer — and to name the specific
+  font family to install.
+
 ## v17.7.0
 
 - The Docker images now run as a non-root user (`app`, uid/gid 1000) by default

pyproject.toml

@@ -6,7 +6,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "ocrmypdf"
-version = "17.7.0"
+version = "17.7.1"
 description = "OCRmyPDF adds an OCR text layer to scanned PDF files, allowing them to be searched"
 readme = "README.md"
 license = "MPL-2.0"
@@ -16,7 +16,7 @@ dependencies = [
   "fpdf2>=2.8.0",
   "img2pdf>=0.5",
   "packaging>=20",
-  "pdfminer.six>=20220319",
+  "pdfminer.six>=20260107",  # fixes parsing of tokens split across the read buffer/streams (gh #1361)
   "pi-heif",                # Heif image format - maintainers: if this is removed, it will NOT break
   "pikepdf>=10",
   "Pillow>=10.0.1",

src/ocrmypdf/_version.py

@@ -1,3 +1,3 @@
 # SPDX-FileCopyrightText: 2022 James R. Barlow
 # SPDX-License-Identifier: MPL-2.0
-__version__ = "17.7.0"
+__version__ = "17.7.1"

src/ocrmypdf/font/multi_font_manager.py

@@ -54,13 +54,15 @@ class MultiFontManager:
         'kok': 'NotoSansDevanagari-Regular',  # Konkani
         'bho': 'NotoSansDevanagari-Regular',  # Bhojpuri
         'mai': 'NotoSansDevanagari-Regular',  # Maithili
-        # CJK
-        'chi': 'NotoSansCJK-Regular',  # Chinese (generic)
-        'zho': 'NotoSansCJK-Regular',  # Chinese (ISO 639-3)
-        'chi_sim': 'NotoSansCJK-Regular',  # Chinese Simplified (Tesseract)
-        'chi_tra': 'NotoSansCJK-Regular',  # Chinese Traditional (Tesseract)
-        'jpn': 'NotoSansCJK-Regular',  # Japanese
-        'kor': 'NotoSansCJK-Regular',  # Korean
+        # CJK — prefer the family matching the document language, because the
+        # modern per-language Noto fonts are region subsets (e.g. NotoSansSC
+        # lacks Japanese kana). The pan-CJK super font is a shared fallback.
+        'chi': 'NotoSansSC-Regular',  # Chinese (generic → Simplified)
+        'zho': 'NotoSansSC-Regular',  # Chinese (ISO 639-3)
+        'chi_sim': 'NotoSansSC-Regular',  # Chinese Simplified (Tesseract)
+        'chi_tra': 'NotoSansTC-Regular',  # Chinese Traditional (Tesseract)
+        'jpn': 'NotoSansJP-Regular',  # Japanese
+        'kor': 'NotoSansKR-Regular',  # Korean
         # Thai
         'tha': 'NotoSansThai-Regular',  # Thai
         # Hebrew
@@ -113,7 +115,14 @@ class MultiFontManager:
         'NotoSans-Regular',  # Latin, Greek, Cyrillic
         'NotoSansArabic-Regular',
         'NotoSansDevanagari-Regular',
+        # Pan-CJK super font first (full coverage), then the per-language
+        # subsets so a glyph missing from one CJK family is found in another.
         'NotoSansCJK-Regular',
+        'NotoSansSC-Regular',
+        'NotoSansTC-Regular',
+        'NotoSansHK-Regular',
+        'NotoSansJP-Regular',
+        'NotoSansKR-Regular',
         'NotoSansThai-Regular',
         'NotoSansHebrew-Regular',
         'NotoSansBengali-Regular',
@@ -256,19 +265,23 @@ class MultiFontManager:
         self._warned_scripts.add(warn_key)
 
         if line_language and line_language in self.LANGUAGE_FONT_MAP:
-            font_name = self.LANGUAGE_FONT_MAP[line_language]
+            font_family = self.LANGUAGE_FONT_MAP[line_language].removesuffix('-Regular')
             log.warning(
-                "No font found with glyphs for '%s' text. "
-                "Install %s for better rendering. "
-                "See https://fonts.google.com/noto",
+                "No installed font has glyphs for the detected '%s' text, so "
+                "it was added as an invisible text layer: it stays searchable "
+                "and copyable, but appears blank when highlighted in a PDF "
+                "viewer. Install the %s font family (via your OS package "
+                "manager or https://fonts.google.com/noto) for full rendering.",
                 line_language,
-                font_name,
+                font_family,
             )
         else:
             log.warning(
-                "No font found with glyphs for some text. "
-                "Install Noto fonts for better rendering. "
-                "See https://fonts.google.com/noto"
+                "No installed font has glyphs for some of the detected text, "
+                "so it was added as an invisible text layer: it stays "
+                "searchable and copyable, but appears blank when highlighted "
+                "in a PDF viewer. Install the matching Noto fonts "
+                "(https://fonts.google.com/noto) for full rendering."
             )
 
     def _has_all_glyphs(self, font: FontManager, text: str) -> bool:

src/ocrmypdf/font/system_font_provider.py

@@ -9,6 +9,7 @@ Linux, macOS, and Windows platforms.
 
 from __future__ import annotations
 
+import glob
 import logging
 import os
 import sys
@@ -75,6 +76,35 @@ class SystemFontProvider:
             # Variable fonts
             'NotoSansCJKsc-VF.otf',
         ],
+        # Per-language CJK families. Modern Google Fonts / Homebrew ship these
+        # as region subset variable fonts ('NotoSansJP[wght].ttf'), matched by
+        # the flexible base search; the legacy per-region super OTFs (full
+        # coverage) are listed here so they also satisfy the logical name.
+        'NotoSansSC-Regular': [
+            'NotoSansSC-Regular.otf',
+            'NotoSansSC-Regular.ttf',
+            'NotoSansCJKsc-Regular.otf',
+        ],
+        'NotoSansTC-Regular': [
+            'NotoSansTC-Regular.otf',
+            'NotoSansTC-Regular.ttf',
+            'NotoSansCJKtc-Regular.otf',
+        ],
+        'NotoSansHK-Regular': [
+            'NotoSansHK-Regular.otf',
+            'NotoSansHK-Regular.ttf',
+            'NotoSansCJKhk-Regular.otf',
+        ],
+        'NotoSansJP-Regular': [
+            'NotoSansJP-Regular.otf',
+            'NotoSansJP-Regular.ttf',
+            'NotoSansCJKjp-Regular.otf',
+        ],
+        'NotoSansKR-Regular': [
+            'NotoSansKR-Regular.otf',
+            'NotoSansKR-Regular.ttf',
+            'NotoSansCJKkr-Regular.otf',
+        ],
         'NotoSansThai-Regular': [
             'NotoSansThai-Regular.ttf',
             'NotoSansThai-Regular.otf',
@@ -149,6 +179,28 @@ class SystemFontProvider:
         ],
     }
 
+    # Font file extensions we know how to load.
+    _FONT_EXTENSIONS = ('.ttf', '.otf', '.ttc')
+
+    # Acceptable filename variants for a font family, ranked best-first.
+    # Lower rank wins when multiple variants of the same family are present.
+    _VARIANT_RANK = {'regular': 0, 'variable': 1, 'vf': 2, 'plain': 3}
+
+    # Extra family bases that can satisfy a logical font, tried after its own
+    # base (so the listed order is the preference). CJK is the case that needs
+    # this: the legacy Adobe-style 'NotoSansCJKsc-Regular.otf' is handled by
+    # NOTO_FONT_PATTERNS, but Homebrew casks and current Google Fonts ship the
+    # per-language families as variable fonts (e.g. 'NotoSansSC[wght].ttf').
+    _ALTERNATE_BASES: dict[str, list[str]] = {
+        'NotoSansCJK-Regular': [
+            'NotoSansSC',  # Simplified Chinese
+            'NotoSansTC',  # Traditional Chinese
+            'NotoSansHK',  # Hong Kong
+            'NotoSansJP',  # Japanese
+            'NotoSansKR',  # Korean
+        ],
+    }
+
     def __init__(self) -> None:
         """Initialize system font provider with empty caches."""
         # Cache: font_name -> FontManager (successfully loaded fonts)
@@ -230,6 +282,76 @@ class SystemFontProvider:
                     # Skip directories we can't read
                     continue
 
+        # No exact static '-Regular' file. Many distributors (Homebrew casks,
+        # current Google Fonts releases) ship Noto fonts as variable fonts with
+        # bracketed axis filenames such as 'NotoSansArabic[wdth,wght].ttf'.
+        # Fall back to a flexible search that also accepts those. See #1652.
+        return self._find_variant_font_file(font_name)
+
+    @staticmethod
+    def _classify_variant(stem: str, base: str) -> str | None:
+        """Classify a font filename stem as a usable variant of ``base``.
+
+        Args:
+            stem: Filename without extension (e.g. 'NotoSansArabic[wdth,wght]')
+            base: Family base name (e.g. 'NotoSansArabic')
+
+        Returns:
+            The variant kind ('regular', 'variable', 'vf', 'plain') or None if
+            the stem is not an acceptable representative of the family. The
+            boundary after ``base`` is required so that 'NotoSans' does not
+            match 'NotoSansArabic', and 'NotoSansArabicUI'/'NotoSansArabic-Bold'
+            do not match a request for 'NotoSansArabic'.
+        """
+        if stem == f'{base}-Regular':
+            return 'regular'
+        if stem.startswith(f'{base}['):  # variable font, e.g. Base[wdth,wght]
+            return 'variable'
+        if stem == f'{base}-VF':  # alternate variable-font naming
+            return 'vf'
+        if stem == base:  # bare family name
+            return 'plain'
+        return None
+
+    def _find_variant_font_file(self, font_name: str) -> Path | None:
+        """Search for a variable font or other acceptable filename variant.
+
+        Tries the font's own family base first, then any alternate bases (used
+        for the modern per-language CJK families). Within that, a static Regular
+        is preferred over a variable font. See issue #1652.
+
+        Args:
+            font_name: Logical font name (e.g. 'NotoSansArabic-Regular')
+
+        Returns:
+            Path to the best-ranked matching font file, or None.
+        """
+        bases = [font_name.removesuffix('-Regular')]
+        bases.extend(self._ALTERNATE_BASES.get(font_name, []))
+
+        # Selection key (base_index, variant_rank): earlier base wins, then the
+        # better variant. Path is carried along but not part of the comparison.
+        best: tuple[tuple[int, int], Path] | None = None
+        for base_index, base in enumerate(bases):
+            for font_dir in self._get_font_dirs():
+                if not font_dir.exists():
+                    continue
+                try:
+                    for path in font_dir.rglob(glob.escape(base) + '*'):
+                        if path.suffix.lower() not in self._FONT_EXTENSIONS:
+                            continue
+                        kind = self._classify_variant(path.stem, base)
+                        if kind is None:
+                            continue
+                        key = (base_index, self._VARIANT_RANK[kind])
+                        if best is None or key < best[0]:
+                            best = (key, path)
+                except PermissionError:
+                    # Skip directories we can't read
+                    continue
+        if best is not None:
+            log.debug("Found system font %s at %s (variant match)", font_name, best[1])
+            return best[1]
         return None
 
     def get_font(self, font_name: str) -> FontManager | None:

src/ocrmypdf/pdfinfo/layout.py

@@ -17,7 +17,6 @@ import pdfminer
 import pdfminer.encodingdb
 import pdfminer.pdfdevice
 import pdfminer.pdfinterp
-import pdfminer.psparser
 from deprecation import deprecated
 from pdfminer.converter import PDFLayoutAnalyzer
 from pdfminer.layout import LAParams, LTChar, LTPage, LTTextBox
@@ -60,12 +59,6 @@ def pdfsimplefont__init__(
 
 PDFSimpleFont.__init__ = pdfsimplefont__init__
 
-# Patch pdfminer.six buffer size
-# The parser doesn't properly handle keyword tokens are split across the end of the
-# buffer, so increase the buffer size something far larger than will ever be seen.
-pdfminer.psparser.PSBaseParser.BUFSIZ = 256 * 1024 * 1024
-
-
 def pdftype3font__pscript5_get_height(self):
     """Monkeypatch for PScript5.dll PDFs.
 

tests/test_multi_font_manager.py

@@ -149,7 +149,9 @@ def test_select_font_for_chinese_language(multi_font_manager):
     if not has_cjk_font(multi_font_manager):
         pytest.skip("CJK font not available")
     font_manager = multi_font_manager.select_font_for_word("你好", "zho")
-    assert font_manager == multi_font_manager.fonts['NotoSansCJK-Regular']
+    # A real, glyph-covering CJK font is selected (which specific family
+    # depends on what is installed: pan-CJK super font or a per-language subset).
+    assert font_manager.font_path.name != 'Occulta.ttf'
 
 
 def test_select_font_for_chinese_generic(multi_font_manager):
@@ -157,7 +159,9 @@ def test_select_font_for_chinese_generic(multi_font_manager):
     if not has_cjk_font(multi_font_manager):
         pytest.skip("CJK font not available")
     font_manager = multi_font_manager.select_font_for_word("中文", "chi")
-    assert font_manager == multi_font_manager.fonts['NotoSansCJK-Regular']
+    # A real, glyph-covering CJK font is selected (which specific family
+    # depends on what is installed: pan-CJK super font or a per-language subset).
+    assert font_manager.font_path.name != 'Occulta.ttf'
 
 
 def test_select_font_for_chinese_simplified(multi_font_manager):
@@ -165,7 +169,9 @@ def test_select_font_for_chinese_simplified(multi_font_manager):
     if not has_cjk_font(multi_font_manager):
         pytest.skip("CJK font not available")
     font_manager = multi_font_manager.select_font_for_word("简体字", "chi_sim")
-    assert font_manager == multi_font_manager.fonts['NotoSansCJK-Regular']
+    # A real, glyph-covering CJK font is selected (which specific family
+    # depends on what is installed: pan-CJK super font or a per-language subset).
+    assert font_manager.font_path.name != 'Occulta.ttf'
 
 
 def test_select_font_for_chinese_traditional(multi_font_manager):
@@ -173,7 +179,9 @@ def test_select_font_for_chinese_traditional(multi_font_manager):
     if not has_cjk_font(multi_font_manager):
         pytest.skip("CJK font not available")
     font_manager = multi_font_manager.select_font_for_word("漢字", "chi_tra")
-    assert font_manager == multi_font_manager.fonts['NotoSansCJK-Regular']
+    # A real, glyph-covering CJK font is selected (which specific family
+    # depends on what is installed: pan-CJK super font or a per-language subset).
+    assert font_manager.font_path.name != 'Occulta.ttf'
 
 
 def test_select_font_for_japanese_language(multi_font_manager):
@@ -181,7 +189,9 @@ def test_select_font_for_japanese_language(multi_font_manager):
     if not has_cjk_font(multi_font_manager):
         pytest.skip("CJK font not available")
     font_manager = multi_font_manager.select_font_for_word("こんにちは", "jpn")
-    assert font_manager == multi_font_manager.fonts['NotoSansCJK-Regular']
+    # A real, glyph-covering CJK font is selected (which specific family
+    # depends on what is installed: pan-CJK super font or a per-language subset).
+    assert font_manager.font_path.name != 'Occulta.ttf'
 
 
 def test_select_font_for_korean_language(multi_font_manager):
@@ -189,7 +199,9 @@ def test_select_font_for_korean_language(multi_font_manager):
     if not has_cjk_font(multi_font_manager):
         pytest.skip("CJK font not available")
     font_manager = multi_font_manager.select_font_for_word("안녕하세요", "kor")
-    assert font_manager == multi_font_manager.fonts['NotoSansCJK-Regular']
+    # A real, glyph-covering CJK font is selected (which specific family
+    # depends on what is installed: pan-CJK super font or a per-language subset).
+    assert font_manager.font_path.name != 'Occulta.ttf'
 
 
 # --- Latin/English Tests ---
@@ -232,7 +244,9 @@ def test_cjk_text_without_language_hint(multi_font_manager):
     if not has_cjk_font(multi_font_manager):
         pytest.skip("CJK font not available")
     font_manager = multi_font_manager.select_font_for_word("你好", None)
-    assert font_manager == multi_font_manager.fonts['NotoSansCJK-Regular']
+    # A real, glyph-covering CJK font is selected (which specific family
+    # depends on what is installed: pan-CJK super font or a per-language subset).
+    assert font_manager.font_path.name != 'Occulta.ttf'
 
 
 def test_fallback_to_occulta_font(multi_font_manager):
@@ -444,3 +458,116 @@ def test_builtin_font_provider_missing_occulta_raises(tmp_path):
     """Test that missing Occulta.ttf raises FileNotFoundError."""
     with pytest.raises(FileNotFoundError, match="Required fallback font"):
         BuiltinFontProvider(tmp_path)
+
+
+class _StubHbFont:
+    """Minimal uharfbuzz Font stand-in with controllable glyph coverage."""
+
+    def __init__(self, covered_codepoints: set[int]):
+        self._covered = covered_codepoints
+
+    def get_nominal_glyph(self, codepoint: int) -> int:
+        return 1 if codepoint in self._covered else 0
+
+
+class _FakeFontManager:
+    """FontManager stand-in whose glyph coverage is fixed per test."""
+
+    def __init__(self, name: str, covered_chars: str):
+        self.font_path = Path(name)
+        self._hb = _StubHbFont({ord(c) for c in covered_chars})
+
+    def get_hb_font(self) -> _StubHbFont:
+        return self._hb
+
+
+class _FakeFontProvider:
+    """FontProvider returning controlled fonts by logical name."""
+
+    def __init__(self, fonts: dict[str, _FakeFontManager]):
+        self._fonts = fonts
+        self._fallback = _FakeFontManager('Occulta.ttf', '')
+
+    def get_font(self, name: str) -> _FakeFontManager | None:
+        return self._fonts.get(name)
+
+    def get_available_fonts(self) -> list[str]:
+        return list(self._fonts)
+
+    def get_fallback_font(self) -> _FakeFontManager:
+        return self._fallback
+
+
+def test_japanese_prefers_jp_family_over_other_cjk():
+    """A Japanese language hint selects NotoSansJP, not another CJK family."""
+    fonts = {
+        'NotoSansSC-Regular': _FakeFontManager('NotoSansSC.ttf', '中'),
+        'NotoSansJP-Regular': _FakeFontManager('NotoSansJP.ttf', '中こ'),
+    }
+    manager = MultiFontManager(font_provider=_FakeFontProvider(fonts))
+    # 'こ' (kana) is only covered by JP; both cover the kanji '中'.
+    font = manager.select_font_for_word('中こ', 'jpn')
+    assert font.font_path.name == 'NotoSansJP.ttf'
+
+
+def test_chinese_simplified_prefers_sc_family():
+    """A Simplified Chinese hint selects NotoSansSC over the pan-CJK font."""
+    fonts = {
+        'NotoSansSC-Regular': _FakeFontManager('NotoSansSC.ttf', '简'),
+        'NotoSansCJK-Regular': _FakeFontManager('NotoSansCJK.ttc', '简'),
+    }
+    manager = MultiFontManager(font_provider=_FakeFontProvider(fonts))
+    font = manager.select_font_for_word('简', 'chi_sim')
+    assert font.font_path.name == 'NotoSansSC.ttf'
+
+
+def test_cjk_fails_over_when_preferred_subset_lacks_glyph():
+    """If the language's subset font lacks a glyph, another CJK family is used."""
+    fonts = {
+        # Simplified Chinese subset cannot render Japanese kana.
+        'NotoSansSC-Regular': _FakeFontManager('NotoSansSC.ttf', '中'),
+        'NotoSansJP-Regular': _FakeFontManager('NotoSansJP.ttf', '中こ'),
+    }
+    manager = MultiFontManager(font_provider=_FakeFontProvider(fonts))
+    # Tagged Simplified Chinese, but the text needs kana only JP covers.
+    font = manager.select_font_for_word('こ', 'chi_sim')
+    assert font.font_path.name == 'NotoSansJP.ttf'
+
+
+def test_cjk_falls_back_to_pan_cjk_super_font():
+    """When only the full-coverage pan-CJK font exists, it serves any CJK lang."""
+    fonts = {
+        'NotoSansCJK-Regular': _FakeFontManager('NotoSansCJK.ttc', '中こ안'),
+    }
+    manager = MultiFontManager(font_provider=_FakeFontProvider(fonts))
+    assert manager.select_font_for_word('こ', 'jpn').font_path.name == 'NotoSansCJK.ttc'
+
+
+def test_missing_cjk_font_warning_names_language_font(font_dir, caplog):
+    """The missing-font warning names the language-specific CJK family (#1652)."""
+    manager = MultiFontManager(font_provider=BuiltinFontProvider(font_dir))
+    with caplog.at_level(logging.WARNING):
+        manager.select_font_for_word("こんにちは", "jpn")
+    assert 'NotoSansJP' in caplog.text
+
+
+def test_missing_font_warning_explains_consequences(font_dir, caplog):
+    """The glyphless-fallback warning should be actionable, not cryptic (#1652).
+
+    With only builtin fonts available, Arabic text cannot be covered, so the
+    manager falls back to glyphless Occulta and must warn helpfully.
+    """
+    # Builtin-only provider: NotoSansArabic is never available, forcing fallback.
+    manager = MultiFontManager(font_provider=BuiltinFontProvider(font_dir))
+
+    with caplog.at_level(logging.WARNING):
+        manager.select_font_for_word("سلام", "fas")
+
+    msg = caplog.text
+    # Identifies the affected language and the font family to install.
+    assert 'fas' in msg
+    assert 'NotoSansArabic' in msg
+    # Explains the user-visible consequence so the message is not cryptic:
+    # the text stays searchable but renders blank when highlighted.
+    assert 'searchable' in msg.lower()
+    assert 'highlight' in msg.lower() or 'select' in msg.lower()

tests/test_system_font_provider.py

@@ -190,6 +190,9 @@ class TestSystemFontProviderAvailableFonts:
         assert 'NotoSansCJK-Regular' in fonts
         assert 'NotoSansArabic-Regular' in fonts
         assert 'NotoSansThai-Regular' in fonts
+        # Per-language CJK families (modern Google Fonts / Homebrew naming)
+        assert 'NotoSansSC-Regular' in fonts
+        assert 'NotoSansJP-Regular' in fonts
 
     def test_fallback_font_raises(self):
         """Test that get_fallback_font raises NotImplementedError."""
@@ -198,6 +201,166 @@ class TestSystemFontProviderAvailableFonts:
             provider.get_fallback_font()
 
 
+class TestSystemFontProviderVariableFonts:
+    """Test discovery of variable fonts and non-static filename variants.
+
+    Homebrew casks and Google Fonts ship Noto fonts as variable fonts with
+    bracketed axis filenames (e.g. ``NotoSansArabic[wdth,wght].ttf``) rather
+    than the static ``NotoSansArabic-Regular.ttf``. See issue #1652.
+    """
+
+    @pytest.fixture
+    def real_font_bytes(self):
+        """Bytes of a real, loadable font (content is irrelevant to the test)."""
+        font_path = (
+            Path(__file__).parent.parent
+            / "src"
+            / "ocrmypdf"
+            / "data"
+            / "NotoSans-Regular.ttf"
+        )
+        if not font_path.exists():
+            pytest.skip("Builtin font not available")
+        return font_path.read_bytes()
+
+    def _provider_for(self, tmp_path, filenames, real_font_bytes):
+        """Build a provider whose only font dir is tmp_path with given files."""
+        for name in filenames:
+            (tmp_path / name).write_bytes(real_font_bytes)
+        provider = SystemFontProvider()
+        provider._font_dirs = [tmp_path]
+        return provider
+
+    def test_finds_variable_font_with_axes(self, tmp_path, real_font_bytes):
+        """A bracketed variable font satisfies a request for the static name."""
+        provider = self._provider_for(
+            tmp_path, ['NotoSansArabic[wdth,wght].ttf'], real_font_bytes
+        )
+        font = provider.get_font('NotoSansArabic-Regular')
+        assert font is not None
+        assert font.font_path.name == 'NotoSansArabic[wdth,wght].ttf'
+
+    def test_finds_weight_only_variable_font(self, tmp_path, real_font_bytes):
+        """A variable font with only a weight axis is also discovered."""
+        provider = self._provider_for(
+            tmp_path, ['NotoSansHebrew[wght].ttf'], real_font_bytes
+        )
+        assert provider.get_font('NotoSansHebrew-Regular') is not None
+
+    def test_variable_font_does_not_cross_match_other_script(
+        self, tmp_path, real_font_bytes
+    ):
+        """The generic NotoSans request must not match a script-specific font."""
+        provider = self._provider_for(
+            tmp_path, ['NotoSansArabic[wdth,wght].ttf'], real_font_bytes
+        )
+        # NotoSans (Latin) must NOT be satisfied by NotoSansArabic.
+        assert provider.get_font('NotoSans-Regular') is None
+
+    def test_does_not_match_ui_or_bold_variants(self, tmp_path, real_font_bytes):
+        """Width/UI and weight variants must not satisfy the Regular request."""
+        provider = self._provider_for(
+            tmp_path,
+            ['NotoSansArabicUI-Regular.ttf', 'NotoSansArabic-Bold.ttf'],
+            real_font_bytes,
+        )
+        assert provider.get_font('NotoSansArabic-Regular') is None
+
+    def test_prefers_static_regular_over_variable(self, tmp_path, real_font_bytes):
+        """When both exist, the static Regular is preferred for predictability."""
+        provider = self._provider_for(
+            tmp_path,
+            ['NotoSansArabic[wdth,wght].ttf', 'NotoSansArabic-Regular.ttf'],
+            real_font_bytes,
+        )
+        font = provider.get_font('NotoSansArabic-Regular')
+        assert font is not None
+        assert font.font_path.name == 'NotoSansArabic-Regular.ttf'
+
+    # --- Modern per-language CJK families (NotoSansSC/TC/HK/JP/KR) ---
+    # Homebrew casks (font-noto-sans-sc, ...) and Google Fonts ship CJK as
+    # variable fonts under these bases rather than the legacy NotoSansCJK*.
+
+    @pytest.mark.parametrize(
+        'filename',
+        [
+            'NotoSansSC[wght].ttf',  # Simplified Chinese (Homebrew/Google)
+            'NotoSansTC[wght].ttf',  # Traditional Chinese
+            'NotoSansHK[wght].ttf',  # Hong Kong
+            'NotoSansJP[wght].ttf',  # Japanese
+            'NotoSansKR[wght].ttf',  # Korean
+        ],
+    )
+    def test_finds_modern_cjk_variable_font(self, tmp_path, real_font_bytes, filename):
+        """A modern per-language CJK variable font satisfies NotoSansCJK."""
+        provider = self._provider_for(tmp_path, [filename], real_font_bytes)
+        font = provider.get_font('NotoSansCJK-Regular')
+        assert font is not None
+        assert font.font_path.name == filename
+
+    def test_finds_static_cjk_language_variant(self, tmp_path, real_font_bytes):
+        """A static per-language CJK Regular also satisfies NotoSansCJK."""
+        provider = self._provider_for(
+            tmp_path, ['NotoSansTC-Regular.otf'], real_font_bytes
+        )
+        assert provider.get_font('NotoSansCJK-Regular') is not None
+
+    def test_prefers_pan_cjk_over_language_variant(self, tmp_path, real_font_bytes):
+        """The pan-CJK family is preferred over a single-language variant."""
+        provider = self._provider_for(
+            tmp_path,
+            ['NotoSansSC[wght].ttf', 'NotoSansCJK[wght].ttf'],
+            real_font_bytes,
+        )
+        font = provider.get_font('NotoSansCJK-Regular')
+        assert font is not None
+        assert font.font_path.name == 'NotoSansCJK[wght].ttf'
+
+    def test_modern_cjk_does_not_cross_match_latin(self, tmp_path, real_font_bytes):
+        """A CJK variable font must not satisfy the generic NotoSans request."""
+        provider = self._provider_for(
+            tmp_path, ['NotoSansSC[wght].ttf'], real_font_bytes
+        )
+        assert provider.get_font('NotoSans-Regular') is None
+
+    # --- Per-language CJK families reachable by their own logical name ---
+    # Needed so MultiFontManager can prefer the family matching the document
+    # language (NotoSansJP for Japanese, NotoSansSC for Simplified Chinese, ...).
+
+    @pytest.mark.parametrize(
+        'logical,filename',
+        [
+            ('NotoSansSC-Regular', 'NotoSansSC[wght].ttf'),
+            ('NotoSansTC-Regular', 'NotoSansTC[wght].ttf'),
+            ('NotoSansHK-Regular', 'NotoSansHK[wght].ttf'),
+            ('NotoSansJP-Regular', 'NotoSansJP[wght].ttf'),
+            ('NotoSansKR-Regular', 'NotoSansKR[wght].ttf'),
+        ],
+    )
+    def test_per_language_cjk_logical_name_resolves(
+        self, tmp_path, real_font_bytes, logical, filename
+    ):
+        """Each per-language CJK family is reachable by its own logical name."""
+        provider = self._provider_for(tmp_path, [filename], real_font_bytes)
+        font = provider.get_font(logical)
+        assert font is not None
+        assert font.font_path.name == filename
+
+    def test_per_language_cjk_static_resolves(self, tmp_path, real_font_bytes):
+        """A static per-language Regular also resolves by logical name."""
+        provider = self._provider_for(
+            tmp_path, ['NotoSansJP-Regular.otf'], real_font_bytes
+        )
+        assert provider.get_font('NotoSansJP-Regular') is not None
+
+    def test_per_language_cjk_does_not_cross_match(self, tmp_path, real_font_bytes):
+        """A JP font must not satisfy an SC request (distinct families)."""
+        provider = self._provider_for(
+            tmp_path, ['NotoSansJP[wght].ttf'], real_font_bytes
+        )
+        assert provider.get_font('NotoSansSC-Regular') is None
+
+
 # --- ChainedFontProvider Tests ---
 
 

uv.lock

@@ -1434,7 +1434,7 @@ wheels = [
 
 [[package]]
 name = "ocrmypdf"
-version = "17.6.0"
+version = "17.7.0"
 source = { editable = "." }
 dependencies = [
     { name = "deprecation" },