v4.5.5 release notes

The 4th argument of re.sub() is maximum number of substitutions,
not flags.

Moreover, re.MULTILINE affects only semantics of ^ and $,
so it wouldn't have any effect on this regular expression.

.gitignore

@@ -4,6 +4,9 @@
 venv*/
 pyvenv.cfg
 tasks.py
+.bash_history
+.ruffus_history.sqlite
+.idea/
 
 # Package building
 *.egg-info/
@@ -27,8 +30,11 @@ htmlcov/
 # Testing
 log/
 /*.pdf
+/*.qdf
+*.ipynb
 .ipynb_checkpoints/
 tests/cache/
 tests/output/
 tests/resources/private/
 tmp/
+pdfbox-app*.jar

.travis.yml

@@ -1,4 +1,3 @@
-sudo: required
 dist: trusty
 language: python
 cache:
@@ -6,50 +5,72 @@ cache:
   directories:
   - packages
   - tests/cache
+  - $HOME/Library/Caches/Homebrew
 
-python:
-- '3.4'
-- '3.5'
-- '3.6'
-
-os:
-- linux
+matrix:
+  include:
+    - os: linux
+      sudo: required
+      language: python
+      python: 3.4
+    - os: linux
+      sudo: required
+      language: python
+      python: 3.5
+    - os: linux
+      sudo: required
+      language: python
+      python: 3.6
+    - os: osx
+      osx_image: xcode8
+      language: generic
+      env:
+        - secure: "hsf6MT+n2x3OiDM2fQyJZdV0/PWYmv81LdVqC6cfnHBE/8N3DloJRqQ7WfO14TxhiK9PEC7MpyCj0lSabUHEO7gSH6Vks6I1asoSkt8S9/bSMlhT4hei+pwVpeGEiU5xHVATNjY+D919VC3IFvc3XmjT74h/2SLhaZ+jhEmDggM="
 
 before_cache:
 - rm -f $HOME/.cache/pip/log/debug.log
 
-before_install:
-- sudo add-apt-repository ppa:vshn/ghostscript -y
-- sudo add-apt-repository ppa:heyarje/libav-11 -y
-- sudo apt-get update -qq
-- sudo apt-get install -y ghostscript tesseract-ocr tesseract-ocr-deu tesseract-ocr-eng
-  tesseract-ocr-fra qpdf poppler-utils libavformat56 libavcodec56 libavutil54 libffi-dev
-- pip install --upgrade pip
-- mkdir -p packages
-- "[ -f packages/unpaper_6.1-1.deb ] || wget -q https://dl.dropboxusercontent.com/u/28971240/unpaper_6.1-1.deb
-  -O packages/unpaper_6.1-1.deb"
-- sudo dpkg -i packages/unpaper_6.1-1.deb
+before_install: |
+  if [[ "$TRAVIS_OS_NAME" == "linux" ]]; then
+    bash .travis/linux_before_install.sh
+  elif [[ "$TRAVIS_OS_NAME" == "osx" ]]; then
+    bash .travis/osx_before_install.sh
+  fi
 
 install:
-- pip install .
-- pip install -r requirements.txt -r test_requirements.txt
+- pip3 install .
+- pip3 install -r requirements.txt -r test_requirements.txt
 
 script:
 - mv ocrmypdf dont_import_this_ocrmypdf
 - pytest
 - mv dont_import_this_ocrmypdf ocrmypdf
 
-# See https://www.appneta.com/blog/pypi-deployment-with-travis-ci/ for
-# steps to set up testpypi deploy for untagged builds if desired
-
 deploy:
-  provider: pypi
+  # release for main pypi
+- provider: pypi
   user: ocrmypdf-travis
   password:
-    secure: DTFOmmNL6olA0+yXvp4u9jXZlZeqrJsJ0526jzqf4a3gZ6jnGTq5UI6WzRsslSyoMMfXKtHQebqHM6ogSgCZinyZ3ufHJo8fn9brxbEc2gsiWkbj5o3bGwdWMT1vNNE7XW0VCpw87rZ1EEwjl4FJHFudMlPR1yfU5+uq0k0PACo=
+    secure: "DTFOmmNL6olA0+yXvp4u9jXZlZeqrJsJ0526jzqf4a3gZ6jnGTq5UI6WzRsslSyoMMfXKtHQebqHM6ogSgCZinyZ3ufHJo8fn9brxbEc2gsiWkbj5o3bGwdWMT1vNNE7XW0VCpw87rZ1EEwjl4FJHFudMlPR1yfU5+uq0k0PACo="
   distributions: "sdist bdist_wheel"
   on:
     branch: master
     tags: true
-    condition: $TRAVIS_PYTHON_VERSION = "3.6"
+    condition: $TRAVIS_PYTHON_VERSION == "3.6" && $TRAVIS_OS_NAME == "linux"
   skip_upload_docs: true
+
+  # test pypi
+- provider: pypi
+  server: https://testpypi.python.org/pypi
+  user: ocrmypdf-travis
+  password:
+    secure: "DTFOmmNL6olA0+yXvp4u9jXZlZeqrJsJ0526jzqf4a3gZ6jnGTq5UI6WzRsslSyoMMfXKtHQebqHM6ogSgCZinyZ3ufHJo8fn9brxbEc2gsiWkbj5o3bGwdWMT1vNNE7XW0VCpw87rZ1EEwjl4FJHFudMlPR1yfU5+uq0k0PACo="
+  distributions: "sdist"
+  on:
+    branch: develop
+    tags: false
+    condition: $TRAVIS_OS_NAME == "osx"
+  skip_upload_docs: true
+
+after_deploy:
+- if [[ "$TRAVIS_OS_NAME" == "osx" ]]; then bash .travis/osx_brew.sh ; fi

.travis/autobrew.py

@@ -0,0 +1,74 @@
+from string import Template
+from subprocess import run, PIPE
+import re
+
+recipe_template = Template("""
+class Ocrmypdf < Formula
+  include Language::Python::Virtualenv
+
+  desc "Adds an OCR text layer to scanned PDF files"
+  homepage "https://github.com/jbarlow83/OCRmyPDF"
+  ${ocrmypdf_url}
+  ${ocrmypdf_sha256}
+
+  depends_on :python3
+  depends_on "pkg-config" => :build
+  depends_on "libffi"
+  depends_on "tesseract"
+  depends_on "ghostscript"
+  depends_on "unpaper"
+  depends_on "qpdf"
+
+  # For Pillow source install
+  depends_on "openjpeg"
+  depends_on "freetype"
+  depends_on "libpng"
+
+${resources}
+
+  def install
+    ENV.append ["SETUPTOOLS_SCM_PRETEND_VERSION"], "v${ocrmypdf_version}"
+    ENV.each do |key, value|
+      puts "#{key}:#{value}"
+    end
+    virtualenv_install_with_resources
+  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"
+  end
+end
+""")
+
+def main():
+    p = run(['poet', '--single', 'ocrmypdf'],
+            encoding='utf-8', stdout=PIPE, check=True)
+
+    ocrmypdf_lines = p.stdout.splitlines()
+    ocrmypdf_url = ocrmypdf_lines[1].strip()
+    ocrmypdf_sha256 = ocrmypdf_lines[2].strip()
+
+    ocrmypdf_version = re.search(
+        r'ocrmypdf-(.+)\.tar.*', ocrmypdf_url).group(1)
+    print(f"Autobrewing {ocrmypdf_version}")
+
+    p = run(['poet', '--resources', 'ocrmypdf'],
+            encoding='utf-8', stdout=PIPE, check=True)
+
+    poet_resources = p.stdout
+
+    # Remove the duplicate "ocrmypdf" resource block
+    all_resources = poet_resources.split('resource')
+    kept_resources = [block for block in all_resources if 'ocrmypdf' not in block]
+    resources = 'resource'.join(kept_resources)
+
+    with open('ocrmypdf.rb', 'w') as out:
+        out.write(recipe_template.substitute(**locals()))
+
+
+if __name__ == '__main__':
+    main()

.travis/linux_before_install.sh

@@ -0,0 +1,24 @@
+#!/bin/bash
+set -euo pipefail
+set -x
+
+sudo add-apt-repository ppa:vshn/ghostscript -y
+sudo add-apt-repository ppa:heyarje/libav-11 -y
+sudo apt-get update -qq
+sudo apt-get install -y \
+	ghostscript \
+	tesseract-ocr \
+	tesseract-ocr-deu \
+	tesseract-ocr-eng \
+	tesseract-ocr-fra \
+	qpdf \
+	poppler-utils \
+	libavformat56 \
+	libavcodec56 \
+	libavutil54 \
+	libffi-dev
+
+pip install --upgrade pip
+mkdir -p packages
+[ -f packages/unpaper_6.1-1.deb ] || wget -q https://dl.dropboxusercontent.com/u/28971240/unpaper_6.1-1.deb -O packages/unpaper_6.1-1.deb
+sudo dpkg -i packages/unpaper_6.1-1.deb

.travis/osx_before_install.sh

@@ -0,0 +1,16 @@
+#!/bin/bash
+set -euo pipefail
+set -x
+
+brew update
+
+brew install openjpeg jbig2dec libtiff     # image libraries
+brew install qpdf
+brew install ghostscript
+brew install python3
+brew install libxml2 libffi leptonica
+brew install unpaper   # optional
+brew install tesseract
+
+pip3 install --upgrade pip
+pip3 install wheel

.travis/osx_brew.sh

@@ -0,0 +1,20 @@
+#!/bin/bash
+set -euo pipefail
+set -x
+
+pip3 install homebrew-pypi-poet
+python3 .travis/autobrew.py
+cat ocrmypdf.rb
+brew audit ocrmypdf.rb
+
+# Important: disable debug output so token is hidden
+set +x
+git clone https://$HOMEBREW_OCRMYPDF_TOKEN@github.com/jbarlow83/homebrew-ocrmypdf.git
+set -x
+
+pushd homebrew-ocrmypdf
+cp ../ocrmypdf.rb Formula/ocrmypdf.rb
+git add Formula/ocrmypdf.rb
+git commit -m "homebrew-ocrmypdf: automatic release $TRAVIS_BUILD_NUMBER $TRAVIS_TAG"
+git push origin master
+popd

Dockerfile.tess4

@@ -1,20 +1,63 @@
 # OCRmyPDF
 #
 # VERSION               4.4.2
-FROM      jbarlow83/ocrmypdf:latest
+FROM 	ubuntu:16.10
 MAINTAINER James R. Barlow <jim@purplerock.ca>
 
-USER root
+RUN apt-get update && apt-get install -y --no-install-recommends \
+  software-properties-common python-software-properties \
+  python3-wheel \
+  python3-reportlab \
+  python3-venv \
+  ghostscript \
+  qpdf \
+  poppler-utils \
+  unpaper \
+  libffi-dev
 
 RUN add-apt-repository ppa:alex-p/tesseract-ocr
 
-RUN apt-get update && apt-get install -y --no-install-recommends \
-	tesseract-ocr-all
+RUN apt-get update \
+	&& apt-get autoremove -y \
+	&& apt-get install -y --no-install-recommends \
+		tesseract-ocr \
+		tesseract-ocr-eng \
+		tesseract-ocr-fra \
+		tesseract-ocr-deu \
+		tesseract-ocr-spa \
+		tesseract-ocr-por \
+		tesseract-ocr-ara \
+		tesseract-ocr-rus \
+		tesseract-ocr-chi-sim 
 
-RUN apt-get autoremove -y && apt-get clean -y
+RUN python3 -m venv --system-site-packages /appenv
+
+# This installs the latest binary wheel instead of the code in the current
+# folder. Installing from source will fail, apparently because cffi needs
+# build-essentials (gcc) to do a source installation 
+# (i.e. "pip install ."). It's unclear to me why this is the case.
+RUN . /appenv/bin/activate; \
+  pip install --upgrade pip \
+  && pip install ocrmypdf
+
+# Now copy the application in, mainly to get the test suite.
+# Do this now to make the best use of Docker cache.
+COPY . /application
+RUN . /appenv/bin/activate; \
+  pip install -r /application/test_requirements.txt
+
+# Remove the junk, including the source version of application since it was
+# already installed
+RUN rm -rf /tmp/* /var/tmp/* /root/* /application/ocrmypdf \
+  && apt-get autoremove -y \
+  && apt-get autoclean -y
+
+RUN useradd docker \
+  && mkdir /home/docker \
+  && chown docker:docker /home/docker
 
 USER docker
-
+WORKDIR /home/docker
 
 # Must use array form of ENTRYPOINT
 # Non-array form does not append other arguments, because that is "intuitive"

README.rst

@@ -35,7 +35,7 @@ Main features
 -  Supports more than `100 languages <https://github.com/tesseract-ocr/tessdata>`_ recognized by Tesseract
 -  Battle-tested on thousands of PDFs, a test suite and continuous integration
 
-For details: please consult the `release notes <RELEASE_NOTES.rst>`_.
+For details: please consult the `documentation <https://ocrmypdf.readthedocs.io/en/latest/>`_.
 
 Motivation
 ----------
@@ -60,7 +60,17 @@ Installation
 Linux, UNIX, and macOS are supported. Windows is not directly supported but there is a Docker image available that runs on Windows.
 
 Users of Debian 9 or later or Ubuntu 16.10 or later may simply
-``apt-get install ocrmypdf``.
+
+.. code-block:: bash
+
+   apt-get install ocrmypdf
+
+and macOS users may simply
+
+.. code-block:: bash
+
+   brew tap jbarlow83/ocrmypdf
+   brew install ocrmypdf
 
 For everyone else, `see our documentation <https://ocrmypdf.readthedocs.io/en/latest/installation.html>`_ for installation steps.
 

docs/batch.rst

@@ -10,11 +10,29 @@ Consider using the excellent `GNU Parallel <https://www.gnu.org/software/paralle
 
 Both ``parallel`` and ``ocrmypdf`` will try to use all available processors. To maximize parallelism without overloading your system with processes, consider using ``parallel -j 2`` to limit parallel to running two jobs at once.
 
-This command will run all ocrmypdf all files named ``*.pdf`` in the current directory and write them to the previous created ``output/`` folder.
+This command will run all ocrmypdf all files named ``*.pdf`` in the current directory and write them to the previous created ``output/`` folder. It will not search subdirectories.
+
+The ``--tag`` argument tells parallel to print the filename as a prefix whenever a message is printed, so that one can trace any errors to the file that produced them.
 
 .. code-block:: bash
 
-	parallel -j 2 ocrmypdf '{}' 'output/{}' ::: *.pdf
+	parallel --tag -j 2 ocrmypdf '{}' 'output/{}' ::: *.pdf
+
+Directory trees
+---------------
+
+This will walk through a directory tree and run OCR on all files in place, printing the output in a way that makes 
+
+.. code-block:: bash
+
+	find . --printf '%p' -name '*.pdf' -exec ocrmypdf '{}' '{}' \;
+
+This only runs one ``ocrmypdf`` process at a time. This variation uses ``find`` to create a directory list and ``parallel`` to parallelize runs of ``ocrmypdf``, again updating files in place.
+
+.. code-block:: bash
+
+	find . -name '*.pdf' | parallel --tag -j 2 ocrmypdf '{}' '{}'
+
 
 Sample script
 """""""""""""

docs/index.rst

@@ -11,14 +11,18 @@ be searched.
 
 PDFs are the best format for scanned documents.  Unfortunately, PDFs can be difficult to work with. OCRmyPDF makes it easy to apply image processing and OCR to existing PDFs.
 
-Contents:
-
 .. toctree::
-   :maxdepth: 2
+   :maxdepth: 1
 
    introduction
+   release_notes
    installation
    languages
+
+.. toctree::
+   :caption: Usage
+   :maxdepth: 2
+
    cookbook
    advanced
    batch
@@ -26,7 +30,6 @@ Contents:
    errors
 
 
-
 Indices and tables
 ==================
 

docs/installation.rst

@@ -10,6 +10,18 @@ Users of Debian 9 ("stretch") or later or Ubuntu 16.10 or later may simply
 
 	apt-get install ocrmypdf
 
+
+Installing on macOS
+-------------------
+
+.. code-block:: bash
+
+   brew tap jbarlow83/ocrmypdf
+   brew install ocrmypdf
+
+
+Users who previously installed OCRmyPDF on macOS may need to remove the ``pip`` based installation (``pip3 uninstall ocrmypdf``).
+
 .. _Docker:
 
 Installing the Docker image
@@ -46,7 +58,8 @@ Assuming you have a Docker engine running, you can download one of the three ava
 +-----------------------------+---------------------------------------------+---------------------------------------------------------------------------------+
 | ocrmypdf-polyglot           | ``docker pull jbarlow83/ocrmypdf-polyglot`` | As above, with all available language packs.                                    |
 +-----------------------------+---------------------------------------------+---------------------------------------------------------------------------------+
-| ocrmypdf-tess4              | ``docker pull jbarlow83/ocrmypdf-tess4``    | Latest ocrmypdf with Tesseract 4.00.00alpha and all language packs.             |
+| ocrmypdf-tess4              | ``docker pull jbarlow83/ocrmypdf-tess4``    | Latest ocrmypdf with Tesseract 4.00.00alpha and English, French, German,        |
+|                             |                                             | Spanish, Portuguese, Chinese Simplified, Arabic and Russian (the top 8).        |
 +-----------------------------+---------------------------------------------+---------------------------------------------------------------------------------+
 
 For example:
@@ -102,8 +115,8 @@ Or in the wonderful `fish shell <https://fishshell.com/>`_:
    funcsave ocrmypdf
 
 
-Installing on macOS
--------------------
+Manual installation on macOS
+----------------------------
 
 These instructions probably work on all macOS supported by Homebrew.
 
@@ -174,10 +187,6 @@ No package is currently available for Ubuntu 16.04, but you can install the depe
       tesseract-ocr \
       qpdf \
       python3-pip \
-      python3-pil \
-      python3-pytest \
-      python3-reportlab \
-      python3-ruffus \
       python3-cffi
 
 If you wish install OCRmyPDF to the system Python, then install as follows (note this installs new packages
@@ -192,7 +201,7 @@ follow these steps.
 
 .. code-block:: bash
 
-   python3 -m venv --system-site-packages venv-ocrmypdf
+   python3 -m venv venv-ocrmypdf
    source venv-ocrmypdf/bin/activate
    pip3 install ocrmypdf
 

docs/introduction.rst

@@ -53,6 +53,20 @@ By default, OCRmyPDF will convert the file to a PDF/A.  This behavior can be dis
 Depending on the settings selected, OCRmyPDF may "graft" the OCR layer into the existing PDF, or reconstruct a visually equivalent new PDF.
 
 
+Why you shouldn't do this manually
+----------------------------------
+
+There are two routes to manually applying OCR to an existing PDF, both of which destroy information in the original PDF.
+
+1. Rasterize each page as an image, OCR the images, and combine the output into a PDF. This preserves the appearance of each page, but resamples all images (possibly losing quality, increasing file size, introducing compression artifacts, etc.)
+
+2. Extract each image, OCR, and combine the output into a PDF. This loses the context in which images are used in the PDF, meaning that cropping, rotation and scaling of pages may be lost. Some PDFs use multiple images per page with stencil masks, which would quite difficult to reassemble correctly. This also loses and text or vector art on any pages in a PDF with both scanned and pure digital content.
+
+In the case of a PDF that is nothing other than a container of images (no rotation, scaling, cropping, one image per page), the second approach is can be lossless.
+
+OCRmyPDF uses several strategies depending on input options and the input PDF itself, but generally speaking it rasterizes a page for OCR and then grafts the OCR back onto the original. As such it can handle complex PDFs and still preserve their contents as much as possible.
+
+
 Limitations
 -----------
 
@@ -73,4 +87,15 @@ Ghostscript also imposes some limitations:
 
 * PDFs containing JBIG2-encoded content will be converted to CCITT Group4 encoding, which has lower compression ratios, if Ghostscript PDF/A is enabled.
   
-OCRmyPDF is currently not designed to be used as a Python API; it is designed to be run as a command line tool. ``import ocrmypf`` currently attempts to process the command line on ``sys.argv`` at import time so it has side effects that will interfere with its use as a package. The API it presents should not be considered stable.
+OCRmyPDF is currently not designed to be used as a Python API; it is designed to be run as a command line tool. ``import ocrmypf`` currently attempts to process the command line on ``sys.argv`` at import time so it has side effects that will interfere with its use as a package. The API it presents should not be considered stable.
+
+
+Similar programs
+----------------
+
+To the author's knowledge, OCRmyPDF is the most feature-rich and thoroughly tested command line OCR PDF conversion tool. If it doesn't meet your needs, consider of these similar open source programs:
+
+* pdf2pdfocr
+* pdfsandwich
+* pypdfocr
+* pdfbeads

docs/release_notes.rst

@@ -1,11 +1,56 @@
-RELEASE NOTES
+Release notes
 =============
 
 OCRmyPDF uses `semantic versioning <http://semver.org/>`_.
 
 
+v4.5.5
+------
+
+-   Automated update of macOS homebrew tap
+-   Fixed issue #154, KeyError '/Contents' when searching for text on blank pages that have no /Contents record
+
+
+v4.5.4
+------
+
+-   Fix ``--skip-big`` raising an exception if a page contains no images (#152) (thanks to @TomRaz)
+-   Fix an issue where pages with no images might trigger "cannot write mode P as JPEG" (#151)
+
+
+v4.5.3
+------
+
+-  Added a workaround for Ghostscript 9.21 and probably earlier versions would fail with the error message "VMerror -25", due to a Ghostscript bug in XMP metadata handling
+-  High Unicode characters (U+10000 and up) are no longer accepted for setting metadata on the command line, as Ghostscript may not handle them correctly.
+-  Fixed an issue where the ``tess4`` renderer would duplicate content onto output pages if tesseract failed or timed out
+-  Fixed ``tess4`` renderer not recognized when lossless reconstruction is possible
+
+
+v4.5.2
+------
+
+-  Fix issue #147. ``--pdf-renderer tess4 --clean`` will produce an oversized page containing the original image in the bottom left corner, due to loss DPI information.
+-  Make "using Tesseract 4.0" warning less ominous
+-  Set up machinery for homebrew OCRmyPDF tap
+
+
+v4.5.1
+------
+
+-  Fix issue #137, proportions of images with a non-square pixel aspect ratio would be distorted in output for ``--force-ocr`` and some other combinations of flags
+
+
+v4.5
+----
+
+-  Exotic PDFs containing "Form XObjects" are now supported (issue #134; PDF reference manual 8.10), and images they contain are taken into account when determining the resolution for rasterizing
+-  The Tesseract 4 Docker image no longer includes all languages, because it took so long to build something would tend to fail
+-  OCRmyPDF now warns about using ``--pdf-renderer tesseract`` with Tesseract 3.04 or lower due to issues with Ghostscript corrupting the OCR text in these cases
+
+
 v4.4.2
-======
+------
 
 -  The Docker images (ocrmypdf, ocrmypdf-polyglot, ocrmypdf-tess4) are now based on Ubuntu 16.10 instead of Debian stretch
 
@@ -16,7 +61,7 @@ v4.4.2
 
 
 v4.4.1
-======
+------
 
 -  To prevent a `TIFF output error <https://github.com/python-pillow/Pillow/issues/2206>`_ caused by img2pdf >= 0.2.1 and Pillow <= 3.4.2, dependencies have been tightened
 -  The Tesseract 4.00 simultaneous process limit was increased from 1 to 2, since it was observed that 1 lowers performance
@@ -25,7 +70,7 @@ v4.4.1
 -  Tweaks to setup.py to deal with issues in the v4.4 release
 
 v4.4
-====
+----
 
 -  Tesseract 4.00 is now supported on an experimental basis.
 
@@ -43,32 +88,32 @@ v4.4
 
 
 v4.3.5
-======
+------
 
 -  Update documentation to confirm Python 3.6.0 compatibility. No code changes were needed, so many earlier versions are likely supported.
 
 
 v4.3.4
-======
+------
 
 -  Fixed "decimal.InvalidOperation: quantize result has too many digits" for high DPI images
 
 
 v4.3.3
-======
+------
 
 -  Fixed PDF/A creation with Ghostscript 9.20 properly
 -  Fixed an exception on inline stencil masks with a missing optional parameter
 
 
 v4.3.2
-======
+------
 
 -  Fixed a PDF/A creation issue with Ghostscript 9.20 (note: this fix did not actually work)
 
 
 v4.3.1
-======
+------
 
 -  Fixed an issue where pages produced by the "hocr" renderer after a Tesseract timeout would be rotated incorrectly if the input page was rotated with a /Rotate marker
 -  Fixed a file handle leak in LeptonicaErrorTrap that would cause a "too many open files" error for files around hundred pages of pages long when ``--deskew`` or ``--remove-background`` or other Leptonica based image processing features were in use, depending on the system value of ``ulimit -n``
@@ -79,7 +124,7 @@ v4.3.1
 
 
 v4.3
-====
+----
 
 -  New feature ``--remove-background`` to detect and erase the background of color and grayscale images
 -  Better documentation
@@ -90,20 +135,20 @@ v4.3
    +  Some output validation is disabled in this mode
 
 v4.2.5
-======
+------
 
 -  Fixed an issue (#100) with PDFs that omit the optional /BitsPerComponent parameter on images
 -  Removed non-free file milk.pdf
 
 
 v4.2.4
-======
+------
 
 -  Fixed an error (#90) caused by PDFs that use stencil masks properly
 -  Fixed handling of PDFs that try to draw images or stencil masks without properly setting up the graphics state (such images are now ignored for the purposes of calculating DPI)
 
 v4.2.3
-======
+------
 
 -  Fixed an issue with PDFs that store page rotation (/Rotate) in an indirect object
 -  Integrated a few fixes to simplify downstream packaging (Debian)
@@ -117,20 +162,20 @@ v4.2.3
 
 
 v4.2.2
-======
+------
 
 -  Improvements to documentation
 
 
 v4.2.1
-======
+------
 
 -  Fixed an issue where PDF pages that contained stencil masks would report an incorrect DPI and cause Ghostscript to abort
 -  Implemented stdin streaming
 
 
 v4.2
-====
+----
 
 -  ocrmypdf will now try to convert single image files to PDFs if they are provided as input (#15)
 
@@ -163,13 +208,13 @@ v4.2
 -  Ghostscript now runs in "safer" mode where possible
 
 v4.1.4
-======
+------
 
 -  Bug fix: monochrome images with an ICC profile attached were incorrectly converted to full color images if lossless reconstruction was not possible due to other settings; consequence was increased file size for these images
 
 
 v4.1.3
-======
+------
 
 -  More helpful error message for PDFs with version 4 security handler
 -  Update usage instructions for Windows/Docker users
@@ -178,14 +223,14 @@ v4.1.3
 
 
 v4.1.2
-======
+------
 
 -  Replace IEC sRGB ICC profile with Debian's sRGB (from icc-profiles-free) which is more compatible with the MIT license
 -  More helpful error message for an error related to certain types of malformed PDFs
 
 
 v4.1
-====
+----
 
 -  ``--rotate-pages`` now only rotates pages when reasonably confidence in the orientation. This behavior can be adjusted with the new argument ``--rotate-pages-threshold``
 -  Fixed problems in error checking if ``unpaper`` is uninstalled or missing at run-time
@@ -193,20 +238,20 @@ v4.1
 
 
 v4.0.7
-======
+------
 
 -  Minor correction to Ghostscript output settings
 
 
 v4.0.6
-======
+------
 
 -  Update install instructions
 -  Provide a sRGB profile instead of using Ghostscript's
 
 
 v4.0.5
-======
+------
 
 -  Remove some verbose debug messages from v4.0.4
 -  Fixed temporary that wasn't being deleted
@@ -214,22 +259,22 @@ v4.0.5
 -  Inline images are now checked during DPI calculation instead of rejecting the image
 
 v4.0.4
-======
+------
 
 Released with verbose debug message turned on. Do not use. Skip to v4.0.5.
 
 
 v4.0.3
-======
+------
 
 New features
-------------
+^^^^^^^^^^^^
 
 -  Page orientations detected are now reported in a summary comment
 
 
 Fixes
------
+^^^^^
 
 -  Show stack trace if unexpected errors occur
 -  Treat "too few characters" error message from Tesseract as a reason to skip that page rather than
@@ -238,10 +283,10 @@ Fixes
 
 
 v4.0.2
-======
+------
 
 Fixes
------
+^^^^^
 
 -  Fixed compatibility with Tesseract 3.04.01 release, particularly its different way of outputting
    orientation information
@@ -250,19 +295,19 @@ Fixes
 
 
 v4.0.1
-======
+------
 
 Fixes
------
+^^^^^
 
 -  Fixed a KeyError if tesseract fails to find page orientation information
 
 
 v4.0
-====
+----
 
 New features
-------------
+^^^^^^^^^^^^
 
 -  Automatic page rotation (``-r``) is now available. It uses ignores any prior rotation information
    on PDFs and sets rotation based on the dominant orientation of detectable text. This feature is
@@ -272,7 +317,7 @@ New features
 
 
 Fixes
------
+^^^^^
 
 -  Fixed an issue where lossless reconstruction could cause some pages to be appear incorrectly
    if the page was rotated by the user in Acrobat after being scanned (specifically if it a /Rotate tag)
@@ -281,7 +326,7 @@ Fixes
 
 
 Changes
--------
+^^^^^^^
 
 -  Logging output is now much easier to read
 -  ``--deskew`` is now performed by Leptonica instead of unpaper (#25)
@@ -294,20 +339,20 @@ Changes
 
 
 v3.2.1
-======
+------
 
 Changes
--------
+^^^^^^^
 
 -  Fixed issue #47 "convert() got and unexpected keyword argument 'dpi'" by upgrading to img2pdf 0.2
 -  Tweaked the Dockerfiles
 
 
 v3.2
-====
+----
 
 New features
-------------
+^^^^^^^^^^^^
 
 -  Lossless reconstruction: when possible, OCRmyPDF will inject text layers without 
    otherwise manipulating the content and layout of a PDF page. For example, a PDF containing a mix
@@ -319,25 +364,25 @@ New features
    for the polyglots among us. It is much larger.
 
 Changes
--------
+^^^^^^^
 
 -  JPEG transcoding quality is now 95 instead of the default 75. Bigger file sizes for less degradation.
 
 
 
 v3.1.1
-======
+------
 
 Changes
--------
+^^^^^^^
 
 -  Fixed bug that caused incorrect page size and DPI calculations on documents with mixed page sizes
 
 v3.1
-====
+----
 
 Changes
--------
+^^^^^^^
 
 -  Default output format is now PDF/A-2b instead of PDF/A-1b
 -  Python 3.5 and macOS El Capitan are now supported platforms - no changes were
@@ -352,10 +397,10 @@ Changes
 -  Set up Travis CI automatic integration testing
 
 v3.0
-====
+----
 
 New features
-------------
+^^^^^^^^^^^^
 
 -  Easier installation with a Docker container or Python's ``pip`` package manager 
 -  Eliminated many external dependencies, so it's easier to setup
@@ -378,7 +423,7 @@ New features
 -  Multiple images on the same PDF page are now supported
 
 Changes
--------
+^^^^^^^
 
 -  New, robust rewrite in Python 3.4+ with ruffus_ pipelines
 -  Now uses Ghostscript 9.14's improved color conversion model to preserve PDF colors
@@ -416,7 +461,7 @@ Changes
 .. _JHOVE: http://jhove.sourceforge.net/
 
 Release candidates
-------------------
+^^^^^^^^^^^^^^^^^^
 
 -  rc9:
 
@@ -484,12 +529,12 @@ where ``settings.txt`` contains *one argument per line*, for example:
 
 
 Fixes
------
+^^^^^
 
 -  Handling of filenames containing spaces: fixed
 
 Notes and known issues
-----------------------
+^^^^^^^^^^^^^^^^^^^^^^
 
 -  Some dependencies may work with lower versions than tested, so try
    overriding dependencies if they are "in the way" to see if they work.
@@ -502,7 +547,7 @@ Notes and known issues
 
 
 v2.2-stable (2014-09-29)
-========================
+------------------------
 
 OCRmyPDF versions 1 and 2 were implemented as shell scripts. OCRmyPDF 3.0+ is a fork that gradually replaced all shell scripts with Python while maintaining the existing command line arguments. No one is maintaining old versions.
 

docs/security.rst

@@ -1,4 +1,4 @@
-PDF Security Issues
+PDF security issues
 ===================
 
 	OCRmyPDF should only be used on PDFs you trust. It is not designed to protect you against malware. 

ocrmypdf/__main__.py

@@ -46,6 +46,10 @@ def complain(message):
     print(*textwrap.wrap(message), file=sys.stderr)
 
 
+# Hack to help debugger context find /usr/local/bin
+if 'IDE_PROJECT_ROOTS' in os.environ:
+    os.environ['PATH'] = '/usr/local/bin:' + os.environ['PATH']
+
 if tesseract.version() < MINIMUM_TESS_VERSION:
     complain(
         "Please install tesseract {0} or newer "
@@ -53,7 +57,6 @@ if tesseract.version() < MINIMUM_TESS_VERSION:
             MINIMUM_TESS_VERSION, tesseract.version()))
     sys.exit(ExitCode.missing_dependency)
 
-
 # -------------
 # Parser
 
@@ -272,13 +275,18 @@ def check_options_output(options, log):
     if options.pdf_renderer == 'auto':
         options.pdf_renderer = 'hocr'
 
-    if options.pdf_renderer == 'tesseract' and \
-            tesseract.version() < '3.04.01' and \
-            os.environ.get('OCRMYPDF_SHARP_TTF', '') != '1':
-        log.warning(
-            "Your version of tesseract has problems with PDF output."
-            " Some PDF viewers will fail to find searchable text.\n"
-            "--pdf-renderer=tesseract is not recommended.")
+    if options.pdf_renderer in ('tesseract', 'tess4'):
+        if tesseract.version() < '3.05':
+            log.warning(
+                "The setting --pdf-renderer=tesseract is not recommend for "
+                " use with tesseract versions less than 3.05, because it "
+                " produces OCR text that is incompatible with Ghostscript and "
+                " some other software.")
+        elif tesseract.version() == '4.00.00alpha':
+            log.warning(
+                "The setting --pdf-renderer={tesseract,tess4} is not"
+                " recommended for builds of tesseract 4.00.00alpha older than"
+                " February 2017. Make sure you are using a recent build.")
 
     if options.debug_rendering and options.pdf_renderer == 'tesseract':
         log.info(
@@ -286,7 +294,7 @@ def check_options_output(options, log):
             "--pdf-renderer=tesseract.")
 
     lossless_reconstruction = False
-    if options.pdf_renderer == 'hocr':
+    if options.pdf_renderer in ('hocr', 'tess4'):
         if not any((options.deskew, options.clean_final, options.force_ocr,
                    options.remove_background)):
             lossless_reconstruction = True
@@ -339,13 +347,31 @@ def check_options_advanced(options, log):
             "commit 3d9fb3b or later")
 
 
+def check_options_metadata(options, log):
+    import unicodedata
+    metadata = [options.title, options.author, options.keywords,
+                options.subject]
+    for s in (m for m in metadata if m):
+        for c in s:
+            if unicodedata.category(c) == 'Co' or ord(c) >= 0x10000:
+                raise ValueError(
+                    "One of the metadata strings contains "
+                    "an unsupported Unicode character: '{}' (U+{})".format(
+                        c, hex(ord(c))[2:].upper()
+                ))
+
+
 def check_options(options, log):
     try:
         check_options_languages(options, log)
+        check_options_metadata(options, log)
         check_options_output(options, log)
         check_options_preprocessing(options, log)
         check_options_ocr_behavior(options, log)
         check_options_advanced(options, log)
+    except ValueError as e:
+        log.error(e)
+        sys.exit(ExitCode.bad_args)
     except argparse.ArgumentError as e:
         log.error(e)
         sys.exit(ExitCode.bad_args)
@@ -354,9 +380,6 @@ def check_options(options, log):
         sys.exit(ExitCode.missing_dependency)
 
 
-
-
-
 # ----------
 # Logging
 
@@ -397,7 +420,7 @@ def available_cpu_count():
 
 
 def cleanup_ruffus_error_message(msg):
-    msg = re.sub(r'\s+', r' ', msg, re.MULTILINE)
+    msg = re.sub(r'\s+', r' ', msg)
     msg = re.sub(r"\((.+?)\)", r'\1', msg)
     msg = msg.strip()
     return msg
@@ -479,10 +502,58 @@ def traverse_ruffus_exception(e_args, options, log):
             return traverse_ruffus_exception(exc, options, log)
 
 
+def check_closed_streams(options):
+    """Work around Python issue with multiprocessing forking on closed streams
+
+    https://bugs.python.org/issue28326
+
+    Attempting to a fork/exec a new Python process when any of std{in,out,err}
+    are closed or not flushable for some reason may raise an exception.
+    Fix this by opening devnull if the handle seems to be closed.  Do this
+    globally to avoid tracking places all places that fork.
+
+    Seems to be specific to multiprocessing.Process not all Python process
+    forkers.
+
+    The error actually occurs when the stream object is not flushable,
+    but replacing an open stream object that is not flushable with
+    /dev/null is a bad idea since it will create a silent failure.  Replacing
+    a closed handle with /dev/null seems safe.
+
+    """
+
+    if sys.stderr is None:
+        sys.stderr = open(os.devnull, 'w')
+
+    if sys.stdin is None:
+        if options.input_file == '-':
+            print("Trying to read from stdin but stdin seems closed",
+                  file=sys.stderr)
+            return False
+        sys.stdin = open(os.devnull, 'r')
+
+    if sys.stdout is None:
+        if options.output_file == '-':
+            # Can't replace stdout if the user is piping
+            # If this case can even happen, it must be some kind of weird
+            # stream.
+            print(textwrap.dedent("""\
+                Output was set to stdout '-' but the stream attached to
+                stdout does not support the flush() system call.  This
+                will fail."""), file=sys.stderr)
+            return False
+        sys.stdout = open(os.devnull, 'w')
+
+    return True
+
+
 def run_pipeline():
     options = parser.parse_args()
     options.verbose_abbreviated_path = 1
 
+    if not check_closed_streams(options):
+        return ExitCode.bad_args
+
     _log, _log_mutex = proxy_logger.make_shared_logger_and_proxy(
         logging_factory, __name__, [None, options.verbose])
     _log.debug('ocrmypdf ' + VERSION)
@@ -523,9 +594,9 @@ def run_pipeline():
                     file."""))
                 return ExitCode.bad_args
         elif not is_file_writable(options.output_file):
-                _log.error(textwrap.dedent("""\
-                    Cutput file location is not writable."""))
-                return ExitCode.file_access_error
+            _log.error(textwrap.dedent("""\
+                Cutput file location is not writable."""))
+            return ExitCode.file_access_error
 
         manager = JobContextManager()
         manager.register('JobContext', JobContext)

ocrmypdf/exec/ghostscript.py

@@ -2,12 +2,32 @@
 # © 2015 James R. Barlow: github.com/jbarlow83
 
 from tempfile import NamedTemporaryFile
-from subprocess import Popen, PIPE, STDOUT, check_call
+from subprocess import Popen, PIPE, STDOUT, check_call, CalledProcessError, \
+    check_output
 from shutil import copy
+from functools import lru_cache
 from . import get_program
 from ..pdfa import SRGB_ICC_PROFILE
 
 
+@lru_cache(maxsize=1)
+def version():
+    args_gs = [
+        get_program('gs'),
+        '--version'
+    ]
+    try:
+        version = check_output(
+                args_gs, close_fds=True, universal_newlines=True,
+                stderr=STDOUT)
+    except CalledProcessError as e:
+        print("Could not find Ghostscript executable on system PATH.",
+              file=sys.stderr)
+        raise MissingDependencyError from e
+
+    return version.strip()
+
+
 def rasterize_pdf(input_file, output_file, xres, yres, raster_device, log,
                   pageno=1):
     with NamedTemporaryFile(delete=True) as tmp:
@@ -61,7 +81,7 @@ def generate_pdfa(pdf_pages, output_file, log, threads=1):
                   universal_newlines=True)
         stdout, _ = p.communicate()
 
-        if 'error' in stdout:
+        if 'error' in stdout or 'ERROR' in stdout:
             log.error(stdout)
         elif 'overprint mode not set' in stdout:
             # Unless someone is going to print PDF/A documents on a

ocrmypdf/exec/tesseract.py

@@ -11,6 +11,7 @@ from ..helpers import page_number
 from . import get_program
 from collections import namedtuple
 from textwrap import dedent
+import PyPDF2 as pypdf
 
 from subprocess import Popen, PIPE, CalledProcessError, \
     TimeoutExpired, check_output, STDOUT, DEVNULL
@@ -186,6 +187,8 @@ def tesseract_log_output(log, stdout, input_file):
             log.warning(prefix + "unsure about page orientation")
         elif 'error' in line.lower() or 'exception' in line.lower():
             log.error(prefix + line.strip())
+        elif 'warning' in line.lower():
+            log.warning(prefix + line.strip())
         elif 'read_params_file' in line.lower():
             log.error(prefix + line.strip())
         else:
@@ -270,6 +273,24 @@ def generate_hocr(input_file, output_hocr, language: list, engine_mode,
                 f_out.write(line)
 
 
+def use_skip_page(text_only, skip_pdf, output_pdf):
+    if not text_only:
+        os.symlink(skip_pdf, output_pdf)
+        return
+
+    # For text only we must create a blank page with dimensions identical
+    # to the skip page because this is equivalent to a page with no text
+
+    pdf_in = pypdf.PdfFileReader(skip_pdf)
+    page0 = pdf_in.pages[0]
+
+    with open(output_pdf, 'wb') as out:
+        pdf_out = pypdf.PdfFileWriter()
+        w, h = page0.mediaBox.getWidth(), page0.mediaBox.getHeight()
+        pdf_out.addBlankPage(w, h)
+        pdf_out.write(out)
+
+
 def generate_pdf(input_image, skip_pdf, output_pdf, language: list,
                  engine_mode, text_only: bool,
                  tessconfig: list, timeout: float, pagesegmode: int, log):
@@ -307,14 +328,14 @@ def generate_pdf(input_image, skip_pdf, output_pdf, language: list,
             universal_newlines=True, timeout=timeout)
     except TimeoutExpired:
         page_timedout(log, input_image)
-        shutil.copy(skip_pdf, output_pdf)
+        use_skip_page(text_only, skip_pdf, output_pdf)
     except CalledProcessError as e:
         tesseract_log_output(log, e.output, input_image)
         if 'read_params_file: parameter not found' in e.output:
             raise TesseractConfigError() from e
 
         if 'Image too large' in e.output:
-            shutil.copy(skip_pdf, output_pdf)
+            use_skip_page(text_only, skip_pdf, output_pdf)
             return
         raise e from e
     else:

ocrmypdf/exec/unpaper.py

@@ -78,7 +78,8 @@ def run(input_file, output_file, dpi, log, mode_args):
             raise e from e
         else:
             log.debug(stdout)
-            Image.open(output_pnm.name).save(output_file)
+            # unpaper sets dpi to 72
+            Image.open(output_pnm.name).save(output_file, dpi=(dpi, dpi))
 
 
 def deskew(input_file, output_file, dpi, log):

ocrmypdf/pageinfo.py

@@ -89,13 +89,13 @@ def _is_unit_square(shorthand):
     pairwise = zip(values, UNIT_SQUARE)
     return all([isclose(a, b, rel_tol=1e-3) for a, b in pairwise])
 
-RasterSettings = namedtuple('RasterSettings',
+XobjectSettings = namedtuple('XobjectSettings',
     ['name', 'shorthand', 'stack_depth'])
 
 InlineSettings = namedtuple('InlineSettings',
     ['settings', 'shorthand', 'stack_depth'])
 
-ContentsInfo = namedtuple('ContentsInfo', ['raster_settings', 'inline_images'])
+ContentsInfo = namedtuple('ContentsInfo', ['xobject_settings', 'inline_images'])
 
 
 def _normalize_stack(operations):
@@ -114,7 +114,7 @@ def _normalize_stack(operations):
             yield (operands, command)
 
 
-def _interpret_contents(contentstream):
+def _interpret_contents(contentstream, initial_shorthand=UNIT_SQUARE):
     """Interpret the PDF content stream
 
     The stack represents the state of the PDF graphics stack.  We are only
@@ -139,8 +139,8 @@ def _interpret_contents(contentstream):
 
     operations = contentstream.operations
     stack = []
-    ctm = _matrix_from_shorthand(UNIT_SQUARE)
-    image_raster_settings = []
+    ctm = _matrix_from_shorthand(initial_shorthand)
+    xobject_settings = []
     inline_images = []
 
     for n, op in enumerate(_normalize_stack(operations)):
@@ -161,10 +161,10 @@ def _interpret_contents(contentstream):
                 _matrix_from_shorthand(operands), ctm)
         elif command == b'Do':
             image_name = operands[0]
-            raster = RasterSettings(
+            settings = XobjectSettings(
                 name=image_name, shorthand=_shorthand_from_matrix(ctm),
                 stack_depth=len(stack))
-            image_raster_settings.append(raster)
+            xobject_settings.append(settings)
         elif command == b'INLINE IMAGE':
             settings = operands['settings']
             inline = InlineSettings(
@@ -173,7 +173,7 @@ def _interpret_contents(contentstream):
             inline_images.append(inline)
 
     return ContentsInfo(
-        raster_settings=image_raster_settings,
+        xobject_settings=xobject_settings,
         inline_images=inline_images)
 
 
@@ -241,8 +241,8 @@ def _get_dpi(ctm_shorthand, image_size):
     return (dpi_w, dpi_h)
 
 
-def _find_page_inline_images(page, pageinfo, contentsinfo):
-    "Find inline images on the page"
+def _find_inline_images(contentsinfo):
+    "Find inline images in the contentstream"
 
     for n, inline in enumerate(contentsinfo.inline_images):
         image = {}
@@ -272,20 +272,44 @@ def _find_page_inline_images(page, pageinfo, contentsinfo):
         yield image
 
 
-def _find_page_regular_images(page, pageinfo, contentsinfo):
-    "Find images stored in XObject resources"
+def _image_xobjects(container):
+    """Search for all XObject-based images in the container
 
-    try:
-        page['/Resources']['/XObject']
-    except KeyError:
+    Usually the container is a page, but it could also be a Form XObject
+    that contains images. Filter out the Form XObjects which are dealt with
+    elsewhere.
+
+    Generate a sequence of tuples (image, xobj container), where container,
+    where xobj is the name of the object and image is the object itself,
+    since the object does not know its own name.
+
+    """
+
+    if '/Resources' not in container:
         return
-    for xobj in page['/Resources']['/XObject']:
-        # PyPDF2 returns the keys as an iterator
-        pdfimage = page['/Resources']['/XObject'][xobj]
-        if pdfimage['/Subtype'] != '/Image':
-            continue
+    resources = container['/Resources']
+    if '/XObject' not in resources:
+        return
+    for xobj in resources['/XObject']:
+        candidate = resources['/XObject'][xobj]
+        if candidate['/Subtype'] == '/Image':
+            image = candidate
+            yield (image, xobj)
+
+
+def _find_regular_images(container, contentsinfo):
+    """Find images stored in the container's /Resources /XObject
+
+    Usually the container is a page, but it could also be a Form XObject
+    that contains images.
+
+    Generates images with their DPI at time of drawing.
+
+    """
+
+    for pdfimage, xobj in _image_xobjects(container):
         image = {}
-        image['name'] = str(xobj)
+        image['name'] = xobj
         image['width'] = pdfimage['/Width']
         image['height'] = pdfimage['/Height']
         if '/BitsPerComponent' in pdfimage:
@@ -329,12 +353,12 @@ def _find_page_regular_images(page, pageinfo, contentsinfo):
 
         image['dpi_w'] = image['dpi_h'] = 0
 
-        for raster in contentsinfo.raster_settings:
+        for xobj in contentsinfo.xobject_settings:
             # Loop in case the same image is display multiple times on a page
-            if raster.name != image['name']:
+            if xobj.name != image['name']:
                 continue
 
-            if raster.stack_depth == 0 and _is_unit_square(raster.shorthand):
+            if xobj.stack_depth == 0 and _is_unit_square(xobj.shorthand):
                 # At least one PDF in the wild (and test suite) draws an image
                 # when the graphics stack depth is 0, meaning that the image
                 # gets drawn into a square of 1x1 PDF units (or 1/72",
@@ -343,7 +367,7 @@ def _find_page_regular_images(page, pageinfo, contentsinfo):
                 continue
 
             dpi_w, dpi_h = _get_dpi(
-                raster.shorthand, (image['width'], image['height']))
+                xobj.shorthand, (image['width'], image['height']))
 
             # When image is used multiple times take the highest DPI it is
             # rendered at
@@ -358,12 +382,91 @@ def _find_page_regular_images(page, pageinfo, contentsinfo):
         yield image
 
 
-def _find_page_images(page, pageinfo, contentsinfo):
-    yield from _find_page_inline_images(page, pageinfo, contentsinfo)
-    yield from _find_page_regular_images(page, pageinfo, contentsinfo)
+def _find_form_xobject_images(pdf, container, contentsinfo):
+    """Find any images that are in Form XObjects in the container
+
+    The container may be a page, or a parent Form XObject.
+
+    """
+    if '/Resources' not in container:
+        return
+    resources = container['/Resources']
+    if '/XObject' not in resources:
+        return
+    for xobj in resources['/XObject']:
+        candidate = resources['/XObject'][xobj]
+        if candidate['/Subtype'] != '/Form':
+            continue
+
+        form_xobject = candidate
+        for settings in contentsinfo.xobject_settings:
+            if settings.name != xobj:
+                continue
+
+            # Find images once for each time this Form XObject is drawn.
+            # This could be optimized to cache the multiple drawing events
+            # but in practice both Form XObjects and multiple drawing of the
+            # same object are both very rare.
+            ctm_shorthand = settings.shorthand
+            yield from _find_images(pdf, form_xobject, ctm_shorthand)
+
+
+def _find_images(pdf, container, shorthand=None):
+    """Find all individual instances of images drawn in the container
+
+    Usually the container is a page, but it may also be a Form XObject.
+
+    On a typical page images are stored inline or as regular images
+    in an XObject.
+
+    Form XObjects may include inline images, XObject images,
+    and recursively, other Form XObjects; and also vector drawing commands.
+
+    Every instance of an image being drawn somewhere is flattened and
+    treated as a unique image, since if the same image is drawn multiple times
+    on one page it may be drawn at differing resolutions, and our objective
+    is to find the resolution at which the page can be rastered without
+    downsampling.
+
+    """
+
+    if container.get('/Type') == '/Page':
+        # For a /Page the content stream is attached to the page's /Contents
+        page = container
+        contentstream = pypdf.pdf.ContentStream(page.getContents(), pdf)
+        initial_shorthand = shorthand or UNIT_SQUARE
+    elif container.get('/Type') == '/XObject' and \
+            container['/Subtype'] == '/Form':
+        # For a Form XObject that content stream is attached to the XObject
+        contentstream = pypdf.pdf.ContentStream(container, pdf)
+
+        # Set the CTM to the state it was when the "Do" operator was
+        # encountered that is drawing this instance of the Form XObject
+        ctm = _matrix_from_shorthand(shorthand or UNIT_SQUARE)
+
+        # A Form XObject may provide its own matrix to map form space into
+        # user space. Get this if one exists
+        form_matrix = _matrix_from_shorthand(
+                container.get('/Matrix', UNIT_SQUARE))
+
+        # Concatenate form matrix with CTM to ensure CTM is correct for
+        # drawing this instance of the XObject
+        ctm = matrix_mult(form_matrix, ctm)
+        initial_shorthand = _shorthand_from_matrix(ctm)
+    else:
+        return
+
+    contentsinfo = _interpret_contents(contentstream, initial_shorthand)
+
+    yield from _find_inline_images(contentsinfo)
+    yield from _find_regular_images(container, contentsinfo)
+    yield from _find_form_xobject_images(pdf, container, contentsinfo)
 
 
 def _page_has_text(pdf, page):
+    if not '/Contents' in page:
+        return False
+
     # Simple test
     text = page.extractText()
     if text.strip() != '':
@@ -405,15 +508,8 @@ def _pdf_get_pageinfo(infile, pageno: int):
     except KeyError:
         pageinfo['rotate'] = 0
 
-    try:
-        contentstream = pypdf.pdf.ContentStream(page.getContents(), pdf)
-    except AttributeError as e:
-        return pageinfo
-
-    contentsinfo = _interpret_contents(contentstream)
-    pageinfo['images'] = [im for im in _find_page_images(
-                                page, pageinfo, contentsinfo)]
-
+    pageinfo['images'] = [im for im in
+                          _find_images(pdf, page)]
     if pageinfo['images']:
         xres = max(image['dpi_w'] for image in pageinfo['images'])
         yres = max(image['dpi_h'] for image in pageinfo['images'])

ocrmypdf/pdfa.py

@@ -26,11 +26,7 @@ pdfa_def_template = u"""%!
 /ICCProfile ($icc_profile)
 def
 
-[ /Title <$title>
-  /Author <$author>
-  /Subject <$subject>
-  /Keywords <$keywords>
-  /Creator <$creator>
+[$docinfo
   /DOCINFO pdfmark
 
 % Define an ICC profile :
@@ -89,20 +85,30 @@ def encode_text_string(s: str) -> str:
 
 
 def _get_pdfa_def(icc_profile, icc_identifier, pdfmark):
-    pdfmark_utf16 = {k: encode_text_string(v) for k, v in pdfmark.items()}
+    # Ghostscript <= 9.21 has a bug where null entries in DOCINFO might produce
+    # ERROR: VMerror (-25) on closing pdfwrite device.
+    # https://bugs.ghostscript.com/show_bug.cgi?id=697684
+    # Work around this by only adding keys that have a nontrivial value
+    docinfo_keys = ('/Title', '/Author', '/Subject', '/Creator', '/Keywords')
+    docinfo_line_template = '  {key} <{value}>'
+
+    def docinfo_gen():
+        for key in docinfo_keys:
+            if key in pdfmark and pdfmark[key].strip() != '':
+                line = docinfo_line_template.format(
+                    key=key, value=encode_text_string(pdfmark[key]))
+                yield line
+
+    docinfo = '\n'.join(docinfo_gen())
 
     t = Template(pdfa_def_template)
     result = t.substitute(icc_profile=icc_profile,
                           icc_identifier=icc_identifier,
-                          title=pdfmark_utf16.get('/Title', ''),
-                          author=pdfmark_utf16.get('/Author', ''),
-                          subject=pdfmark_utf16.get('/Subject', ''),
-                          creator=pdfmark_utf16.get('/Creator', ''),
-                          keywords=pdfmark_utf16.get('/Keywords', ''))
+                          docinfo=docinfo)
     return result
 
 
-def generate_pdfa_def(target_filename, pdfmark, icc='sRGB'):
+def generate_pdfa_ps(target_filename, pdfmark, icc='sRGB'):
     if icc == 'sRGB':
         icc_profile = SRGB_ICC_PROFILE
     else:
@@ -116,6 +122,10 @@ def generate_pdfa_def(target_filename, pdfmark, icc='sRGB'):
         f.write(ps)
 
 
+# The old name is generate_pdfa_def -- now deprecated
+generate_pdfa_def = generate_pdfa_ps
+
+
 def file_claims_pdfa(filename):
     """Determines if the file claims to be PDF/A compliant
 

ocrmypdf/pipeline.py

@@ -23,7 +23,7 @@ from ruffus import formatter, regex, Pipeline, suffix
 
 from .hocrtransform import HocrTransform
 from .pageinfo import pdf_get_all_pageinfo
-from .pdfa import generate_pdfa_def, file_claims_pdfa
+from .pdfa import generate_pdfa_ps, file_claims_pdfa
 from .helpers import re_symlink, is_iterable_notstr, page_number
 from .exec import ghostscript, tesseract, qpdf
 from .exceptions import *
@@ -247,7 +247,7 @@ def is_ocr_required(pageinfo, log, options):
                                 "skipping all processing on this page"))
             ocr_required = False
 
-    if ocr_required and options.skip_big:
+    if ocr_required and options.skip_big and pageinfo['images']:
         pixel_count = pageinfo['width_pixels'] * pageinfo['height_pixels']
         if pixel_count > (options.skip_big * 1000000):
             ocr_required = False
@@ -393,15 +393,16 @@ def rasterize_with_ghostscript(
     pageinfo = get_pageinfo(input_file, context)
 
     device = 'png16m'  # 24-bit
-    if all(image['comp'] == 1 for image in pageinfo['images']):
-        if all(image['bpc'] == 1 for image in pageinfo['images']):
-            device = 'pngmono'
-        elif all(image['bpc'] > 1 and image['color'] == 'index'
-                 for image in pageinfo['images']):
-            device = 'png256'
-        elif all(image['bpc'] > 1 and image['color'] == 'gray'
-                 for image in pageinfo['images']):
-            device = 'pnggray'
+    if pageinfo['images']:
+        if all(image['comp'] == 1 for image in pageinfo['images']):
+            if all(image['bpc'] == 1 for image in pageinfo['images']):
+                device = 'pngmono'
+            elif all(image['bpc'] > 1 and image['color'] == 'index'
+                     for image in pageinfo['images']):
+                device = 'png256'
+            elif all(image['bpc'] > 1 and image['color'] == 'gray'
+                     for image in pageinfo['images']):
+                device = 'pnggray'
 
     log.debug("Rasterize {0} with {1}".format(
               os.path.basename(input_file), device))
@@ -517,7 +518,8 @@ def select_visible_page_image(
     image = next(ii for ii in infiles if ii.endswith(image_suffix))
 
     pageinfo = get_pageinfo(image, context)
-    if all(orig_image['enc'] == 'jpeg' for orig_image in pageinfo['images']):
+    if pageinfo['images'] and \
+            all(im['enc'] == 'jpeg' for im in pageinfo['images']):
         # If all images were JPEGs originally, produce a JPEG as output
         im = Image.open(image)
 
@@ -555,16 +557,21 @@ def select_image_layer(
         re_symlink(page_pdf, output_file, log)
     else:
         pageinfo = get_pageinfo(image, context)
-        dpi = get_page_dpi(pageinfo, options)
-        dpi = float(dpi[0]), float(dpi[1])
-        layout_fun = img2pdf.get_fixed_dpi_layout_fun(dpi)
+
+        # We rasterize a square DPI version of each page because most image
+        # processing tools don't support rectangular DPI. Use the square DPI
+        # as it accurately describes the image. It would be possible to
+        # resample the image at this stage back to non-square DPI to more
+        # closely resemble the input, except that the hocr renderer does not
+        # understand non-square DPI. The tess4 renderer would be fine.
+        dpi = get_page_square_dpi(pageinfo, options)
+        layout_fun = img2pdf.get_fixed_dpi_layout_fun((dpi, dpi))
 
         with open(image, 'rb') as imfile, \
                 open(output_file, 'wb') as pdf:
-            rawdata = imfile.read()
             log.debug('{:4d}: convert'.format(page_number(page_pdf)))
             img2pdf.convert(
-                rawdata, with_pdfrw=False,
+                imfile, with_pdfrw=False,
                 layout_fun=layout_fun, outputstream=pdf)
             log.debug('{:4d}: convert done'.format(page_number(page_pdf)))
 
@@ -762,13 +769,14 @@ def generate_postscript_stub(
     options = context.get_options()
     pdf = pypdf.PdfFileReader(input_file)
     pdfmark = get_pdfmark(pdf, options)
-    generate_pdfa_def(output_file, pdfmark)
+    generate_pdfa_ps(output_file, pdfmark)
 
 
 def skip_page(
         input_file,
         output_file,
-        log):
+        log,
+        context):
     # The purpose of this step is its filter to forward only the skipped
     # files (.skip.oriented.pdf) while disregarding the processed ones
     # (.ocr.oriented.pdf).  Alternative would be for merge_pages to filter
@@ -1031,7 +1039,7 @@ def build_pipeline(options, work_folder, log, context):
         task_func=generate_postscript_stub,
         input=task_repair_pdf,
         filter=formatter(r'\.repaired\.pdf'),
-        output=os.path.join(work_folder, 'pdfa_def.ps'),
+        output=os.path.join(work_folder, 'pdfa.ps'),
         extras=[log, context])
     task_generate_postscript_stub.active_if(options.output_type == 'pdfa')
 
@@ -1043,7 +1051,7 @@ def build_pipeline(options, work_folder, log, context):
         filter=suffix('.skip.oriented.pdf'),
         output='.done.pdf',
         output_dir=work_folder,
-        extras=[log])
+        extras=[log, context])
 
     # Merge pages
     task_merge_pages_ghostscript = main_pipeline.merge(

tests/conftest.py

@@ -22,10 +22,21 @@ def is_linux():
     return platform.system() == 'Linux'
 
 
+@pytest.helpers.register
+def is_macos():
+    return platform.system() == 'Darwin'
+
+
 @pytest.helpers.register
 def running_in_docker():
-    # Docker creates a file named /.dockerinit
-    return os.path.exists('/.dockerinit')
+    # Docker creates a file named /.dockerenv (newer versions) or
+    # /.dockerinit (older) -- this is undocumented, not an offical test
+    return os.path.exists('/.dockerenv') or os.path.exists('/.dockerinit')
+
+
+@pytest.helpers.register
+def running_in_travis():
+    return os.environ.get('TRAVIS') == 'true'
 
 
 TESTS_ROOT = os.path.abspath(os.path.dirname(__file__))
@@ -90,7 +101,7 @@ def check_ocrmypdf(input_file, output_file, *args, env=None):
 
     p, out, err = run_ocrmypdf(input_file, output_file, *args, env=env)
     #print(err)  # ensure py.test collects the output, use -s to view
-    assert p.returncode == 0
+    assert p.returncode == 0, "<stderr>\n" + err + "\n</stderr>"
     assert os.path.exists(str(output_file)), "Output file not created"
     assert os.stat(str(output_file)).st_size > 100, "PDF too small or empty"
     assert out == "", \
@@ -114,3 +125,11 @@ def run_ocrmypdf(input_file, output_file, *args, env=None):
     #print(err)
 
     return p, out, err
+
+
+@pytest.helpers.register
+def first_page_dimensions(pdf):
+    from ocrmypdf import pageinfo
+    info = pageinfo.pdf_get_all_pageinfo(str(pdf))
+    page0 = info[0]
+    return (page0['width_inches'], page0['height_inches'])

tests/resources/README.rst

@@ -64,6 +64,9 @@ under the terms of the license in LICENSE.rst.
     *   - epson.pdf
         - @lowesjam
         - a linearized PDF containing some unusual indirect objects, created by an Epson printer; printout of a Wikipedia article (CC BY-SA)
+    *   - formxobject.pdf
+        - @jbarlow83
+        - hand-crafted exotic PDF containing an image inside a Form XObject
     *   - francais.pdf
         - @jbarlow83
         - a page containing French accents (diacritics)  

tests/test_main.py

@@ -12,6 +12,7 @@ import PyPDF2 as pypdf
 from ocrmypdf.exceptions import ExitCode
 from ocrmypdf import leptonica
 from ocrmypdf.pdfa import file_claims_pdfa
+from ocrmypdf.exec import ghostscript
 
 
 check_ocrmypdf = pytest.helpers.check_ocrmypdf
@@ -166,48 +167,50 @@ def test_preserve_metadata(spoof_tesseract_noop, output_type,
     assert pdfa_info['output'] == output_type
 
 
-@pytest.mark.skipif(
-    pytest.helpers.is_linux() and not pytest.helpers.running_in_docker(),
-    reason="likely to fail if Linux locale is not configured correctly")
 @pytest.mark.parametrize("output_type", [
     'pdfa', 'pdf'
     ])
 def test_override_metadata(spoof_tesseract_noop, output_type, resources,
                            outpdf):
     input_file = resources / 'c02-22.pdf'
-
     german = 'Du siehst den Wald vor lauter Bäumen nicht.'
     chinese = '孔子'
-    high_unicode = 'U+1030C is: 𐌌'
 
     p, out, err = run_ocrmypdf(
         input_file, outpdf,
         '--title', german,
         '--author', chinese,
-        '--subject', high_unicode,
         '--output-type', output_type,
         env=spoof_tesseract_noop)
 
-    assert p.returncode == ExitCode.ok
+    assert p.returncode == ExitCode.ok, err
 
-    pdf = str(outpdf)
+    reader = pypdf.PdfFileReader(outpdf)
 
-    out_pdfinfo = check_output(['pdfinfo', pdf], universal_newlines=True)
-    lines_pdfinfo = out_pdfinfo.splitlines()
-    pdfinfo = {}
-    for line in lines_pdfinfo:
-        k, v = line.strip().split(':', maxsplit=1)
-        pdfinfo[k.strip()] = v.strip()
+    assert reader.documentInfo['/Title'] == german
+    assert reader.documentInfo['/Author'] == chinese
+    assert reader.documentInfo.get('/Keywords', '') == ''
 
-    assert pdfinfo['Title'] == german
-    assert pdfinfo['Author'] == chinese
-    assert pdfinfo['Subject'] == high_unicode
-    assert pdfinfo.get('Keywords', '') == ''
-
-    pdfa_info = file_claims_pdfa(pdf)
+    pdfa_info = file_claims_pdfa(outpdf)
     assert pdfa_info['output'] == output_type
 
 
+def test_high_unicode(spoof_tesseract_noop, resources, no_outpdf):
+
+    # Ghostscript doesn't support high Unicode, so neither do we, to be
+    # safe
+    input_file = resources / 'c02-22.pdf'
+    high_unicode = 'U+1030C is: 𐌌'
+
+    p, out, err = run_ocrmypdf(
+        input_file, no_outpdf,
+        '--subject', high_unicode,
+        '--output-type', 'pdfa',
+        env=spoof_tesseract_noop)
+
+    assert p.returncode == ExitCode.bad_args, err
+
+
 @pytest.mark.parametrize('renderer', [
     'hocr',
     'tesseract',
@@ -409,10 +412,14 @@ def test_force_ocr_on_pdf_with_no_images(spoof_tesseract_crash, resources,
     assert not os.path.exists(no_outpdf)
 
 
+@pytest.mark.skipif(
+    pytest.helpers.is_macos() and pytest.helpers.running_in_travis(),
+    reason="takes too long to install language packs in Travis macOS homebrew")
 def test_french(spoof_tesseract_cache, resources, outpdf):
     p, out, err = run_ocrmypdf(
         resources / 'francais.pdf', outpdf, '-l', 'fra',
         env=spoof_tesseract_cache)
+    print(os.environ)
     assert p.returncode == ExitCode.ok, \
         "This test may fail if Tesseract language packs are missing"
 
@@ -564,6 +571,46 @@ def test_non_square_resolution(renderer, spoof_tesseract_cache,
     assert in_pageinfo[0]['yres'] == out_pageinfo[0]['yres']
 
 
+@pytest.mark.skipif(sys.version_info < (3, 5), reason="needs math.isclose")
+@pytest.mark.parametrize('renderer', [
+    'hocr',
+    'tesseract'
+    ])
+def test_convert_to_square_resolution(renderer, spoof_tesseract_cache,
+                                      resources, outpdf):
+    from math import isclose
+
+    # Confirm input image is non-square resolution
+    in_pageinfo = pdf_get_all_pageinfo(str(resources / 'aspect.pdf'))
+    assert in_pageinfo[0]['xres'] != in_pageinfo[0]['yres']
+
+    # --force-ocr requires means forced conversion to square resolution
+    check_ocrmypdf(
+        resources / 'aspect.pdf', outpdf,
+        '--force-ocr',
+        '--pdf-renderer', renderer, env=spoof_tesseract_cache)
+
+    out_pageinfo = pdf_get_all_pageinfo(str(outpdf))
+
+    in_p0, out_p0 = in_pageinfo[0], out_pageinfo[0]
+
+    # Resolution show now be equal
+    assert out_p0['xres'] == out_p0['yres']
+
+    # Page size should match input page size
+    assert isclose(in_p0['width_inches'],
+                   out_p0['width_inches'])
+    assert isclose(in_p0['height_inches'],
+                   out_p0['height_inches'])
+
+    # Because we rasterized the page to produce a new image, it should occupy
+    # the entire page
+    out_im_w = out_p0['images'][0]['width'] / out_p0['images'][0]['dpi_w']
+    out_im_h = out_p0['images'][0]['height'] / out_p0['images'][0]['dpi_h']
+    assert isclose(out_p0['width_inches'], out_im_w)
+    assert isclose(out_p0['height_inches'], out_im_h)
+
+
 def test_image_to_pdf(spoof_tesseract_noop, resources, outpdf):
     check_ocrmypdf(
         resources / 'LinnSequencer.jpg', outpdf, '--image-dpi', '200',
@@ -614,6 +661,23 @@ def test_stdout(spoof_tesseract_noop, ocrmypdf_exec, resources, outpdf):
     assert qpdf.check(output_file, log=None)
 
 
+def test_closed_streams(spoof_tesseract_noop, ocrmypdf_exec, resources, outpdf):
+    input_file = str(resources / 'francais.pdf')
+    output_file = str(outpdf)
+
+    def evil_closer():
+        os.close(0)
+        os.close(1)
+
+    p_args = ocrmypdf_exec + [input_file, output_file]
+    p = Popen(
+        p_args, close_fds=True, stdout=None, stderr=PIPE, stdin=None,
+        env=spoof_tesseract_noop, preexec_fn=evil_closer)
+    out, err = p.communicate()
+    print(err.decode())
+    assert p.returncode == ExitCode.ok
+
+
 def test_masks(spoof_tesseract_noop, resources, outpdf):
     check_ocrmypdf(resources / 'masks.pdf', outpdf, env=spoof_tesseract_noop)
 
@@ -742,3 +806,41 @@ THIS FILE IS INVALID
         '--tesseract-config', str(cfg_file))
     assert "parameter not found" in err, "No error message"
     assert p.returncode == ExitCode.invalid_config
+
+
+def test_form_xobject(spoof_tesseract_noop, resources, outpdf):
+    check_ocrmypdf(resources / 'formxobject.pdf', outpdf,
+                   '--force-ocr',
+                   env=spoof_tesseract_noop)
+
+
+@pytest.mark.skipif(sys.version_info < (3, 5), reason="needs math.isclose")
+@pytest.mark.parametrize('renderer', [
+    'hocr',
+    'tesseract',
+])
+def test_pagesize_consistency(renderer, resources, outpdf):
+    from math import isclose
+
+    first_page_dimensions = pytest.helpers.first_page_dimensions
+
+    infile = resources / 'linn.pdf'
+
+    before_dims = first_page_dimensions(infile)
+
+    check_ocrmypdf(
+        infile,
+        outpdf, '--pdf-renderer', renderer,
+        '--clean', '--deskew', '--remove-background', '--clean-final')
+
+    after_dims = first_page_dimensions(outpdf)
+
+    assert isclose(before_dims[0], after_dims[0])
+    assert isclose(before_dims[1], after_dims[1])
+
+
+def test_skip_big_with_no_images(spoof_tesseract_noop, resources, outpdf):
+    check_ocrmypdf(resources / 'blank.pdf', outpdf,
+                   '--skip-big', '5',
+                   '--force-ocr',
+                   env=spoof_tesseract_noop)

tests/test_pageinfo.py

@@ -101,3 +101,10 @@ def test_jpeg(resources, outdir):
     assert pdfimage['enc'] == 'jpeg'
     assert (pdfimage['dpi_w'] - 150) < 1e-5
 
+
+def test_form_xobject(resources):
+    filename = resources / 'formxobject.pdf'
+
+    pdfinfo = pageinfo.pdf_get_all_pageinfo(str(filename))
+    pdfimage = pdfinfo[0]['images'][0]
+    assert pdfimage['width'] == 50

tests/test_tess4.py

@@ -4,19 +4,110 @@
 import pytest
 from ocrmypdf.exceptions import ExitCode
 from ocrmypdf.exec import tesseract
+from ocrmypdf import pageinfo
+import sys
+import os
+import PyPDF2 as pypdf
+
+
+spoof = pytest.helpers.spoof
+
+
+def tess4_possible_location():
+    """The location of tesseract 4 may be OCRMYPDF_TESS4, OCRMYPDF_TESSERACT,
+    or the installed version on PATH."""
+    return os.environ.get('OCRMYPDF_TESS4') or \
+           os.environ.get('OCRMYPDF_TESSERACT') or \
+            'tesseract'
+
+
+@pytest.fixture
+def ensure_tess4():
+    return spoof(tesseract=tess4_possible_location())
+
+
+def tess4_available():
+    """Check if a tesseract 4 binary is available, even if it's not the
+    official "tesseract" on PATH
+
+    """
+    old_environ = os.environ.copy()
+    try:
+        os.environ['OCRMYPDF_TESSERACT'] = tess4_possible_location()
+        return tesseract.v4() and tesseract.has_textonly_pdf()
+    finally:
+        os.environ = old_environ
 
 
 # Skip all tests in this file if not tesseract 4
-pytestmark = pytest.mark.skipif(not tesseract.v4(),
-                                reason="tesseract 4.0 required")
+pytestmark = pytest.mark.skipif(
+    not tess4_available(),
+    reason="tesseract 4.0 with textonly_pdf feature required")
+
+check_ocrmypdf = pytest.helpers.check_ocrmypdf
+run_ocrmypdf = pytest.helpers.run_ocrmypdf
+spoof = pytest.helpers.spoof
 
 
-@pytest.mark.skipif(not tesseract.has_textonly_pdf(),
-                    reason="requires textonly_pdf feature")
-def test_textonly_pdf(resources, outdir):
-    pytest.helpers.check_ocrmypdf(
+def test_textonly_pdf(ensure_tess4, resources, outdir):
+    check_ocrmypdf(
         resources / 'linn.pdf',
-        outdir / 'linn_textonly.pdf', '--pdf-renderer', 'tess4')
+        outdir / 'linn_textonly.pdf', '--pdf-renderer', 'tess4',
+        env=ensure_tess4)
 
 
+@pytest.mark.skipif(sys.version_info < (3, 5), reason="needs math.isclose")
+def test_pagesize_consistency_tess4(ensure_tess4, resources, outpdf):
+    from math import isclose
 
+    infile = resources / 'linn.pdf'
+
+    before_dims = pytest.helpers.first_page_dimensions(infile)
+
+    check_ocrmypdf(
+        infile,
+        outpdf, '--pdf-renderer', 'tess4',
+        '--clean', '--deskew', '--remove-background', '--clean-final',
+        env=ensure_tess4)
+
+    after_dims = pytest.helpers.first_page_dimensions(outpdf)
+
+    assert isclose(before_dims[0], after_dims[0])
+    assert isclose(before_dims[1], after_dims[1])
+
+
+@pytest.mark.parametrize('basename', ['graph_ocred.pdf', 'cardinal.pdf'])
+def test_skip_pages_does_not_replicate(
+        ensure_tess4, resources, basename, outdir):
+    infile = resources / basename
+    outpdf = outdir / basename
+
+    check_ocrmypdf(
+        infile,
+        outpdf, '--pdf-renderer', 'tess4', '--force-ocr',
+        '--tesseract-timeout', '0',
+        env=ensure_tess4
+    )
+
+    info_in = pageinfo.pdf_get_all_pageinfo(str(infile))
+
+    info = pageinfo.pdf_get_all_pageinfo(str(outpdf))
+    for page in info:
+        assert len(page['images']) == 1, "skipped page was replicated"
+
+    for n in range(len(info_in)):
+        assert info[n]['width_inches'] == info_in[n]['width_inches']
+
+
+def test_content_preservation(ensure_tess4, resources, outpdf):
+    infile = resources / 'masks.pdf'
+
+    check_ocrmypdf(
+        infile,
+        outpdf, '--pdf-renderer', 'tess4', '--tesseract-timeout', '0',
+        env=ensure_tess4
+    )
+
+    info = pageinfo.pdf_get_all_pageinfo(str(outpdf))
+    page = info[0]
+    assert len(page['images']) > 1, "masked were rasterized"