Ghostcript 9.21 seems to have a regression related to Unicode metadata

RELEASE_NOTES.rst

@@ -7,7 +7,7 @@ OCRmyPDF uses `semantic versioning <http://semver.org/>`_.
 v4.5.2
 ======
 
--  Fix issue #147, unpaper loses DPI information, which affects PDF rendering with ``--pdf-renderer tess4``
+-  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
 

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:

tests/test_main.py

@@ -12,6 +12,7 @@
 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
@@ -172,6 +173,10 @@ def test_preserve_metadata(spoof_tesseract_noop, output_type,
 @pytest.mark.skipif(
     pytest.helpers.is_macos() and pytest.helpers.running_in_travis(),
     reason="save Travis the trouble of installing poppler")
+@pytest.mark.xfail(
+    ghostscript.version() == '9.21',
+    reason="gs 9.21 has a regression that affects this"
+    )
 @pytest.mark.parametrize("output_type", [
     'pdfa', 'pdf'
     ])
@@ -191,7 +196,7 @@ def test_override_metadata(spoof_tesseract_noop, output_type, resources,
         '--output-type', output_type,
         env=spoof_tesseract_noop)
 
-    assert p.returncode == ExitCode.ok
+    assert p.returncode == ExitCode.ok, err
 
     pdf = str(outpdf)