Ghostscript spoof scripts did not report their --version correctly

This caused trouble before with test_deskew

.travis/autobrew.py

@@ -1,5 +1,5 @@
 #!/usr/bin/env python3
-# © 2016 James R. Barlow: github.com/jbarlow83
+# © 2016-7 James R. Barlow: github.com/jbarlow83
 
 from string import Template
 from subprocess import run, PIPE
@@ -14,43 +14,69 @@ class Ocrmypdf < Formula
   ${ocrmypdf_url}
   ${ocrmypdf_sha256}
 
-  depends_on :python3
-  depends_on :x11  # Pillow needs XQuartz
-  depends_on "pkg-config" => :build
-  depends_on "libffi"
-  depends_on "tesseract"
-  depends_on "ghostscript"
-  depends_on "unpaper"
-  depends_on "qpdf"
-
-  # mactex installs its own ghostscript by default which causes problems
-  # mactex users should use caskroom/cask/mactex-no-ghostscript instead
-  conflicts_with cask: "caskroom/cask/mactex"
-
-  # For Pillow source install
-  depends_on "openjpeg"
   depends_on "freetype"
+  depends_on "ghostscript"
+  depends_on "jpeg"
   depends_on "libpng"
-  depends_on "libjpeg"
-  depends_on "webp"
-  depends_on "little-cms2"
-  depends_on "zlib"
+  depends_on "pkg-config" => :build
+  depends_on "python3"
+  depends_on "qpdf"
+  depends_on "tesseract"
+  depends_on "unpaper"
 
 ${resources}
   def install
-    ENV.append ["SETUPTOOLS_SCM_PRETEND_VERSION"], "v${ocrmypdf_version}"
-    ENV.each do |key, value|
-      puts "#{key}:#{value}"
+    venv = virtualenv_create(libexec, "python3")
+
+    resource("Pillow").stage do
+      inreplace "setup.py" do |s|
+        sdkprefix = MacOS::CLT.installed? ? "" : MacOS.sdk_path
+        s.gsub! "openjpeg.h", "probably_not_a_header_called_this_eh.h"
+        s.gsub! "ZLIB_ROOT = None", "ZLIB_ROOT = ('#{sdkprefix}/usr/lib', '#{sdkprefix}/usr/include')"
+        s.gsub! "JPEG_ROOT = None", "JPEG_ROOT = ('#{Formula["jpeg"].opt_prefix}/lib', '#{Formula["jpeg"].opt_prefix}/include')"
+        s.gsub! "FREETYPE_ROOT = None", "FREETYPE_ROOT = ('#{Formula["freetype"].opt_prefix}/lib', '#{Formula["freetype"].opt_prefix}/include')"
+      end
+
+      # avoid triggering "helpful" distutils code that doesn't recognize Xcode 7 .tbd stubs
+      ENV.append "CFLAGS", "-I#{MacOS.sdk_path}/System/Library/Frameworks/Tk.framework/Versions/8.5/Headers" unless MacOS::CLT.installed?
+      venv.pip_install Pathname.pwd
     end
-    virtualenv_install_with_resources
+
+    res = resources.map(&:name).to_set - ["Pillow"]
+
+    res.each do |r|
+      venv.pip_install resource(r)
+    end
+
+    venv.pip_install_and_link buildpath
   end
 
   test do
-    # `test do` will create, run in and delete a temporary directory.
-    #
-    # The installed folder is not in the path, so use the entire path to any
-    # executables being tested: `system "#{bin}/program", "do", "something"`.
-    system "#{bin}/ocrmypdf", "--version"
+    # Since we use Python 3, we require a UTF-8 locale
+    ENV["LC_ALL"] = "en_US.UTF-8"
+
+    # One page Postscript with the wording "Testing" on the page
+    # This is more compact than including a test PDF
+    (testpath/"test.ps").write(
+      <<~EOS
+        %!PS
+        /Times-Roman findfont
+        20 scalefont
+        setfont
+        gsave
+        newpath
+        200 400 moveto
+        (Testing) show
+        closepath
+        stroke
+        showpage
+      EOS
+    )
+
+    system "#{Formula["ghostscript"].opt_bin}/ps2pdf", testpath/"test.ps", testpath/"test.pdf"
+
+    # Use ocrmypdf -f to rasterize the PDF to image before doing OCR
+    system "#{bin}/ocrmypdf", "-f", "-q", "--deskew", testpath/"test.pdf", testpath/"ocr.pdf"
   end
 end
 """)

docs/cookbook.rst

@@ -157,6 +157,29 @@ If you set ``--tesseract-timeout 0`` OCRmyPDF will apply its image processing wi
     ocrmypdf --tesseract-timeout=0 --remove-background input.pdf output.pdf
 
 
+Redo OCR
+""""""""
+
+To redo OCR on a file OCRed with other OCR software or a previous version of OCRmyPDF and/or Tesseract, you may use the ``--force-ocr`` argument. Normally, OCRmyPDF does not modify files that already appear to contain OCR text.
+
+.. code-block:: bash
+
+    ocrmypdf --force-ocr input.pdf output.pdf
+
+Note that the method above will force rasterization of all pages, potentially reducing quality or losing vector content. 
+
+To ensure quality is preserved, one could extract all of the images and rebuild the PDF for a lossless transformation. This recipe does not work when PDFs contain multiple images per page, as many do in practice. It will also lose any page rotation information.
+
+.. code-block:: bash
+
+    pdfimages -all old-ocr.pdf prefix  # extract all images
+    img2pdf -o temp.pdf prefix*        # construct new PDF from the images
+    # review the new PDF to ensure it visually matches the old one
+    ocrmypdf --output-type pdf temp.pdf new-ocr.pdf
+
+``--output-type pdf`` is used here to avoid using Ghostscript which will also rasterize images.
+
+
 Improving OCR quality
 ---------------------
 

docs/release_notes.rst

@@ -5,6 +5,15 @@ OCRmyPDF uses `semantic versioning <http://semver.org/>`_ for its command line i
 
 The OCRmyPDF package itself does not contain a public API, although it is fairly stable and breaking changes are usually timed with a major release. A future release will clearly define the stable public API.
 
+v5.5
+----
+
+-   Add new argument ``--max-image-mpixels``. Pillow 5.0 now raises an exception when images may be decompression bombs. This argument can be used to override the limit Pillow sets.
+-   Fix output page cropped when using the sandwich renderer and OCR is skipped on a rotated and image-processed page
+-   A warning is now issued when old versions of Ghostscript are used in cases known to cause issues with non-Latin characters
+-   Fix a few parameter validation checks for ``-output-type pdfa-1`` and ``pdfa-2`` 
+
+
 v5.4.4
 ------
 

ocrmypdf/__main__.py

@@ -14,6 +14,7 @@ import logging
 import argparse
 
 import PyPDF2 as pypdf
+import PIL
 
 import ruffus.ruffus_exceptions as ruffus_exceptions
 import ruffus.cmdline as cmdline
@@ -23,7 +24,7 @@ from .pipeline import JobContext, JobContextManager, \
     cleanup_working_files, build_pipeline
 from .pdfa import file_claims_pdfa
 from .helpers import is_iterable_notstr, re_symlink, is_file_writable
-from .exec import tesseract, qpdf
+from .exec import tesseract, qpdf, ghostscript
 from . import PROGRAM_NAME, VERSION
 
 from .exceptions import *
@@ -241,6 +242,11 @@ ocrsettings.add_argument(
 advanced = parser.add_argument_group(
     "Advanced",
     "Advanced options to control Tesseract's OCR behavior")
+advanced.add_argument(
+    '--max-image-mpixels', action='store', type=float, metavar='MPixels',
+    help="Set maximum number of pixels to unpack before treating an image as a "
+         "decompression bomb",
+    default=128.0)
 advanced.add_argument(
     '--tesseract-config', action='append', metavar='CFG', default=[],
     help="Additional Tesseract configuration files -- see documentation")
@@ -321,11 +327,12 @@ def check_options_languages(options, _log):
     if '+' in options.language[0]:
         options.language = options.language[0].split('+')
 
-    if not set(options.language).issubset(tesseract.languages()):
+    languages = set(options.language)
+    if not languages.issubset(tesseract.languages()):
         msg = (
             "The installed version of tesseract does not have language "
             "data for the following requested languages: \n")
-        for lang in (set(options.language) - tesseract.languages()):
+        for lang in (languages - tesseract.languages()):
             msg += lang + '\n'
         raise MissingDependencyError(msg)
 
@@ -348,7 +355,8 @@ def check_options_output(options, log):
         options.pdf_renderer = 'sandwich'
 
     if options.pdf_renderer == 'tesseract':
-        if tesseract.version() < '3.05' and options.output_type == 'pdfa':
+        if tesseract.version() < '3.05' and \
+                options.output_type.startswith('pdfa'):
             log.warning(
                 "For best results use --pdf-renderer=tesseract "
                 "--output-type=pdf to disable PDF/A generation via "
@@ -412,9 +420,9 @@ def check_options_ocr_behavior(options, log):
     # if options.redo_ocr and (options.skip_text or options.force_ocr):
     #     raise argparse.ArgumentError(
     #         "Error: --redo-ocr and other OCR options are incompatible.")
-
+    languages = set(options.language)
     if options.pdf_renderer == 'hocr' and \
-            not set(options.language).issubset(HOCR_OK_LANGS):
+            not languages.issubset(HOCR_OK_LANGS):
         msg = (
             "The 'hocr' PDF renderer is known to cause problems with one "
             "or more of the languages in your document. ")
@@ -427,6 +435,15 @@ def check_options_ocr_behavior(options, log):
                 "Use --pdf-renderer tesseract --output-type pdf to avoid "
                 "this issue")
         log.warning(msg)
+    elif ghostscript.version() < '9.20' and \
+            not languages.issubset(HOCR_OK_LANGS) \
+            and options.output_type != 'pdf':
+        msg = (
+            "The installed version of Ghostscript does not work correctly "
+            "with the OCR languages you specified. Use --output-type pdf or "
+            "upgrade to Ghostscript 9.20 or later to avoid this issue.")
+        msg += "Found Ghostscript {}".format(ghostscript.version())
+        log.warning(msg)
 
 
 def check_options_advanced(options, log):
@@ -441,10 +458,10 @@ def check_options_advanced(options, log):
             "--pdf-renderer tess4 requires Tesseract 4.x "
             "commit 3d9fb3b or later")
     if options.pdfa_image_compression != 'auto' and \
-            options.output_type != 'pdfa':
+            options.output_type.startswith('pdfa'):
         log.warning(
             "--pdfa-image-compression argument has no effect when "
-            "--output-type is not 'pdfa'"
+            "--output-type is not 'pdfa', 'pdfa-1', or 'pdfa-2'"
         )
 
 
@@ -587,6 +604,12 @@ def do_ruffus_exception(ruffus_five_tuple, options, log):
 
             """))
         exit_code = ExitCode.encrypted_pdf
+    elif exc_name == 'PIL.Image.DecompressionBombError':
+        msg = cleanup_ruffus_error_message(exc_value)
+        msg += ("\nUse the --max-image-mpixels argument to set increase the "
+                "maximum number of megapixels to accept.")
+        log.error(msg)
+        exit_code = ExitCode.input_file
 
     if exit_code is not None:
         return exit_code
@@ -686,6 +709,10 @@ def run_pipeline():
 
     check_options(options, _log)
 
+    PIL.Image.MAX_IMAGE_PIXELS = int(options.max_image_mpixels * 1000000)
+    if PIL.Image.MAX_IMAGE_PIXELS == 0:
+        PIL.Image.MAX_IMAGE_PIXELS = None
+
     # Complain about qpdf version < 7.0.0
     # Suppress the warning if in the test suite, since there are no PPAs
     # for qpdf 7.0.0 for Ubuntu trusty (i.e. Travis)

ocrmypdf/exec/__init__.py

@@ -29,8 +29,14 @@ def get_version(program, *,
             stdout=PIPE, stderr=STDOUT, check=True)
         output = proc.stdout
     except CalledProcessError as e:
-        raise MissingDependencyError(
-            "Could not find program '{}' on the PATH".format(program)) from e
+        if get_program(program) == program:
+            raise MissingDependencyError(
+                "Could not find program '{}' on the PATH".format(
+                    program)) from e
+        else:
+            raise MissingDependencyError(
+                "Could not find program '{}'".format(
+                    get_program(program))) from e            
 
     try:
         version = re.match(regex, output.strip()).group(1)

ocrmypdf/exec/ghostscript.py

@@ -37,7 +37,7 @@ def rasterize_pdf(input_file, output_file, xres, yres, raster_device, log,
     :param yres: 
     :param raster_device: 
     :param log: 
-    :param pageno: page number to rasterize
+    :param pageno: page number to rasterize (beginning at page 1)
     :param page_dpi: resolution tuple (x, y) overriding output image DPI 
     :return: 
     """
@@ -55,11 +55,11 @@ def rasterize_pdf(input_file, output_file, xres, yres, raster_device, log,
             '-sDEVICE=%s' % raster_device,
             '-dFirstPage=%i' % pageno,
             '-dLastPage=%i' % pageno,
-            '-o', tmp.name,
             '-r{0}x{1}'.format(str(int_res[0]), str(int_res[1])),
+            '-o', tmp.name,
             fspath(input_file)
         ]
-
+        
         p = run(args_gs, stdout=PIPE, stderr=STDOUT,
                 universal_newlines=True)
         if _gs_error_reported(p.stdout):
@@ -75,6 +75,7 @@ def rasterize_pdf(input_file, output_file, xres, yres, raster_device, log,
         # if the resolution happens to be fractional, then the discrepancy
         # would change the size of the output page, especially if the DPI
         # is quite low. Resize the image to the expected size
+
         tmp.seek(0)
         with Image.open(tmp) as im:
             expected_size = round(im.size[0] / int_res[0] * res[0]), \

ocrmypdf/exec/tesseract.py

@@ -279,6 +279,10 @@ def use_skip_page(text_only, skip_pdf, output_pdf, output_text):
     with open(output_pdf, 'wb') as out:
         pdf_out = pypdf.PdfFileWriter()
         w, h = page0.mediaBox.getWidth(), page0.mediaBox.getHeight()
+        # If skip page has a /Rotate key, replicate the rotation
+        rotation = int(page0.get('/Rotate', 0))
+        if rotation % 180 == 90:
+            w, h = h, w
         pdf_out.addBlankPage(w, h)
         pdf_out.write(out)
 

requirements.txt

@@ -2,7 +2,7 @@
 # setup.py lists a separate set of requirements that are looser to simplify
 # installation
 ruffus == 2.6.3
-Pillow == 4.3.0
+Pillow == 5.0.0
 reportlab == 3.4.0
 PyPDF2 == 1.26.0
 img2pdf == 0.2.4

tests/spoof/gs_feature_elision.py

@@ -22,7 +22,7 @@ not permitted in PDF/A-2, overprint mode not set"""
 def main():
     if '--version' in sys.argv:
         print('9.20')
-        print('SPOOFED: ' + os.path.basename(__filename__))
+        print('SPOOFED: ' + os.path.basename(__file__))
         sys.exit(0)
 
     gs_args = ['gs'] + sys.argv[1:]

tests/spoof/gs_pdfa_failure.py

@@ -17,7 +17,7 @@ def real_ghostscript(argv):
 def main():
     if '--version' in sys.argv:
         print('9.20')
-        print('SPOOFED: ' + os.path.basename(__filename__))
+        print('SPOOFED: ' + os.path.basename(__file__))
         sys.exit(0)
 
     # Unless some argument is calling for PDFA generation, forward to

tests/spoof/gs_raster_failure.py

@@ -16,7 +16,7 @@ def real_ghostscript(argv):
 def main():
     if '--version' in sys.argv:
         print('9.20')
-        print('SPOOFED: ' + os.path.basename(__filename__))
+        print('SPOOFED: ' + os.path.basename(__file__))
         sys.exit(0)
 
     # For any rendering calls (device == pdfwrite) call real ghostscript

tests/spoof/gs_render_failure.py

@@ -16,7 +16,7 @@ def real_ghostscript(argv):
 def main():
     if '--version' in sys.argv:
         print('9.20')
-        print('SPOOFED: ' + os.path.basename(__filename__))
+        print('SPOOFED: ' + os.path.basename(__file__))
         sys.exit(0)
 
     # For any rasterize calls (device != pdfwrite) call real ghostscript

tests/spoof/tesseract_noop.py

@@ -73,10 +73,11 @@ def main():
             output = sys.argv[-3]
             with Image.open(inputf) as im:
                 dpi = im.info['dpi']
-                imsize = im.size[0] * dpi[0] / 72, im.size[1] * dpi[1] / 72
+                pagesize = im.size[0] / dpi[0], im.size[1] / dpi[1]
+                ptsize = pagesize[0] * 72, pagesize[1] * 72
 
             pdf_out = pypdf.PdfFileWriter()
-            pdf_out.addBlankPage(imsize[0], imsize[1])
+            pdf_out.addBlankPage(ptsize[0], ptsize[1])
             with open(output + '.pdf', 'wb') as f:
                 pdf_out.write(f)
             with open(output + '.txt', 'w') as f:

tests/test_main.py

@@ -14,6 +14,8 @@ from ocrmypdf.exec import ghostscript, tesseract, qpdf
 import logging
 from math import isclose
 
+import PIL
+
 
 check_ocrmypdf = pytest.helpers.check_ocrmypdf
 run_ocrmypdf = pytest.helpers.run_ocrmypdf
@@ -69,11 +71,10 @@ def test_quick(spoof_tesseract_cache, resources, outpdf):
     check_ocrmypdf(resources / 'ccitt.pdf', outpdf, env=spoof_tesseract_cache)
 
 
-@pytest.mark.filterwarnings('ignore:Image size')
 def test_deskew(spoof_tesseract_noop, resources, outdir):
     # Run with deskew
     deskewed_pdf = check_ocrmypdf(
-        resources / 'skew.pdf', outdir / 'skew.pdf', '-d', '-v', '1',
+        resources / 'skew.pdf', outdir / 'skew.pdf', '-d',
         env=spoof_tesseract_noop)
 
     # Now render as an image again and use Leptonica to find the skew angle
@@ -88,7 +89,8 @@ def test_deskew(spoof_tesseract_noop, resources, outdir):
         xres=150,
         yres=150,
         raster_device='pngmono',
-        log=log)
+        log=log,
+        pageno=1)
 
     from ocrmypdf.leptonica import Pix
     pix = Pix.read(str(deskewed_png))
@@ -127,8 +129,8 @@ def test_remove_background(spoof_tesseract_noop, resources, outdir):
         xres=100,
         yres=100,
         raster_device='png16m',
-        log=log)
-
+        log=log,
+        pageno=1)
 
     # The output image should contain pure white and black
     im = Image.open(output_png)
@@ -521,7 +523,9 @@ def test_tesseract_image_too_big(renderer, spoof_tesseract_big_image_error,
                                  resources, outpdf):
     check_ocrmypdf(
         resources / 'hugemono.pdf', outpdf, '-r',
-        '--pdf-renderer', renderer, env=spoof_tesseract_big_image_error)
+        '--pdf-renderer', renderer, 
+        '--max-image-mpixels', '0',
+        env=spoof_tesseract_big_image_error)
 
 
 def test_no_unpaper(resources, no_outpdf):
@@ -1037,3 +1041,44 @@ def test_bad_utf8(spoof_tess_bad_utf8, renderer, resources, no_outpdf):
     assert p.returncode != 0
     assert 'not utf-8' in err, "should whine about utf-8"
     assert '\\x96' in err, 'should repeat backslash encoded output'
+
+
+@pytest.mark.skipif(
+    not tesseract.has_textonly_pdf(), 
+    reason="issue only affects sandwich")
+def test_rotate_deskew_timeout(resources, outdir):
+    check_ocrmypdf(
+        resources / 'rotated_skew.pdf',
+        outdir / 'deskewed.pdf',
+        '--deskew',
+        '--tesseract-timeout', '0',
+        '--pdf-renderer', 'sandwich'
+    )
+
+    correlation = check_monochrome_correlation(
+        outdir,
+        reference_pdf=resources / 'ccitt.pdf',
+        reference_pageno=1,
+        test_pdf=outdir / 'deskewed.pdf',
+        test_pageno=1)
+
+    # Confirm that the page still got deskewed
+    assert correlation > 0.50
+
+
+@pytest.mark.skipif(
+    PIL.PILLOW_VERSION < '5.0.0',
+    reason="Pillow < 5.0.0 doesn't raise the exception")
+def test_decompression_bomb(resources, outpdf):
+    p, out, err = run_ocrmypdf(
+        resources / 'hugemono.pdf',
+        outpdf
+    )
+    assert 'decompression bomb' in err
+
+    p, out, err = run_ocrmypdf(
+        resources / 'hugemono.pdf',
+        outpdf,
+        '--max-image-mpixels', '2000'
+    )
+    assert p.returncode == 0