Fix issue when weave handoff occurs with no OCR font present

If using --tesseract-timeout 0 and any image processing on a file with
more than 100 pages, the weave handoff will occur. Ensure this
works correctly even if no Glyphless font is present.

Closes #347

src/ocrmypdf/_weave.py

@@ -17,6 +17,7 @@
 
 from itertools import groupby
 from pathlib import Path
+import os
 
 import pikepdf
 
@@ -24,6 +25,9 @@ from .exec import tesseract
 from .helpers import flatten_groups, page_number
 
 
+MAX_OPEN_PAGE_PDFS = int(os.environ.get('_OCRMYPDF_MAX_OPEN_PAGE_PDFS', 100))
+
+
 def _update_page_resources(*, page, font, font_key, procset):
     """Update this page's fonts with a reference to the Glyphless font"""
 
@@ -34,7 +38,7 @@ def _update_page_resources(*, page, font, font_key, procset):
         fonts = resources['/Font']
     except KeyError:
         fonts = pikepdf.Dictionary({})
-    if font_key not in fonts:
+    if font_key is not None and font_key not in fonts:
         fonts[font_key] = font
     resources['/Font'] = fonts
 
@@ -177,6 +181,8 @@ def _find_font(text, pdf_base):
             break
     if pdf_text_font:
         font = pdf_base.copy_foreign(pdf_text_font)
+    if font_key is None:
+        print('font_key is None')
     return font, font_key
 
 
@@ -246,10 +252,10 @@ def _fix_toc(pdf_base, pageref_remap, log):
         invalidated to its new one.
         """
         try:
-        pageref = dest_node[0]
-        if pageref['/Type'] == '/Page' and pageref.objgen in pageref_remap:
-            new_objgen = pageref_remap[pageref.objgen]
-            dest_node[0] = pdf_base.get_object(new_objgen)
+            pageref = dest_node[0]
+            if pageref['/Type'] == '/Page' and pageref.objgen in pageref_remap:
+                new_objgen = pageref_remap[pageref.objgen]
+                dest_node[0] = pdf_base.get_object(new_objgen)
         except (IndexError, TypeError) as e:
             log.warning("This file may contain invalid table of contents entries")
             log.debug(e)
@@ -392,7 +398,7 @@ def weave_layers(infiles, output_file, log, context):
             content_rotation - autorotate_correction
         ) % 360
 
-        if len(keep_open) > 100:
+        if len(keep_open) > MAX_OPEN_PAGE_PDFS:
             # qpdf limitations require us to keep files open when we intend
             # to copy content from them before saving. However, we want to keep
             # a lid on file handles and memory usage, so for big files we're
@@ -409,7 +415,7 @@ def weave_layers(infiles, output_file, log, context):
 
             pdf_base = pikepdf.open(interim)
             procset = pdf_base.pages[0].Resources.ProcSet
-            font = pdf_base.pages[0].Resources.Font.get(font_key)
+            font, font_key = None, None  # Reacquire this information
 
     _fix_toc(pdf_base, pagerefs, log)
     pdf_base.save(output_file)

src/ocrmypdf/exec/tesseract.py

@@ -349,7 +349,7 @@ def generate_pdf(
         if os.path.exists(prefix + '.txt'):
             shutil.move(prefix + '.txt', output_text)
     except TimeoutExpired:
-        page_timedout(log, input_image)
+        page_timedout(log, input_image, timeout)
         use_skip_page(text_only, skip_pdf, output_pdf, output_text)
     except CalledProcessError as e:
         tesseract_log_output(log, e.output, input_image)

tests/test_weave.py

@@ -17,20 +17,24 @@
 
 from unittest.mock import MagicMock
 import logging
+import os
 
 import pytest
 
 import pikepdf
-from ocrmypdf._weave import _fix_toc
+from ocrmypdf._weave import _fix_toc, _update_page_resources
 
-def test_invalid_toc(resources, tmpdir, caplog):
+check_ocrmypdf = pytest.helpers.check_ocrmypdf
+
+
+def test_invalid_toc(resources, outdir, caplog):
     pdf = pikepdf.open(resources / 'toc.pdf')
 
     # Corrupt a TOC entry
     pdf.Root.Outlines.Last.Dest = pikepdf.Array([None, 0.0, 0.1, 0.2])
-    pdf.save(tmpdir / 'test.pdf')
+    pdf.save(outdir / 'test.pdf')
 
-    pdf = pikepdf.open(tmpdir / 'test.pdf')
+    pdf = pikepdf.open(outdir / 'test.pdf')
     remap = {}
     remap[pdf.pages[0].objgen] = pdf.pages[0].objgen  # Dummy remap
 
@@ -38,3 +42,23 @@ def test_invalid_toc(resources, tmpdir, caplog):
     log = logging.getLogger()
     _fix_toc(pdf, remap, log)
     assert 'invalid table of contents entries' in caplog.text
+
+
+def test_no_glyphless_weave(resources, outdir):
+    pdf = pikepdf.open(resources / 'francais.pdf')
+    pdf_aspect = pikepdf.open(resources / 'aspect.pdf')
+    pdf_cmyk = pikepdf.open(resources / 'cmyk.pdf')
+    pdf.pages.extend(pdf_aspect.pages)
+    pdf.pages.extend(pdf_cmyk.pages)
+    pdf.save(outdir / 'test.pdf')
+
+    env = os.environ.copy()
+    env['_OCRMYPDF_MAX_OPEN_PAGE_PDFS'] = '2'
+    check_ocrmypdf(
+        outdir / 'test.pdf',
+        outdir / 'out.pdf',
+        '--deskew',
+        '--tesseract-timeout',
+        '0',
+        env=env,
+    )