Fix OCR bug in Qubes Fedora 38 templates

Provide a fix for an OCR bug that affected Fedora 38 templates of Qubes
OS. In that specific configuration, the PyMuPDF version accepts the
Tesseract data directory only from the `TESSDATA_PREFIX` environment
variable. Our mistake was that we were setting this environment variable
in a dev script, instead of setting it for all configurations.

In this commit, we set an attribute in the fitz.fitz module, so that
both dev scripts and end-user installations can work. This is hacky, but
it targets an old PyMuPDF release after all, so we don't expect things
to break in the long run.

Fixes #737

dangerzone/conversion/pixels_to_pdf.py

@@ -64,8 +64,32 @@ async def convert(
                         tessdata=get_tessdata_dir(),
                     )
                 else:
-                    # XXX method signature changed in v1.22.5 to add tessdata arg
-                    # TODO remove after oldest distro has PyMuPDF >= v1.22.5
+                    # XXX: In PyMuPDF v1.22.5, the function signature of
+                    # `pdfocr_tobytes()` / `pdfocr_save()` was extended with an argument
+                    # to explicitly set the Tesseract data dir [1].
+                    #
+                    # In earlier versions, the PyMuPDF developers recommend setting this
+                    # path via the TESSDATA_PREFIX environment variable. In practice,
+                    # this environment variable is read at import time, so subsequent
+                    # changes to the environment variable are not tracked [2].
+                    #
+                    # To make things worse, any attempt to alter the internal attribute
+                    # (`fitz.TESSDATA_PREFIX`) makes no difference as well, when using
+                    # the OCR functions. That's due to the way imports work in `fitz`,
+                    # where somehow the internal `fitz.fitz` module is shadowed.
+                    #
+                    # A hacky solution is to grab the `fitz.fitz` module from
+                    # `sys.modules`, and set there the TESSDATA_PREFIX variable. We can
+                    # get away with this hack because we have a proper solution for
+                    # subsequent PyMuPDF versions, and we know that nothing will change
+                    # in older versions.
+                    #
+                    # TODO: Remove after oldest distro has PyMuPDF >= v1.22.5
+                    #
+                    # [1]: https://pymupdf.readthedocs.io/en/latest/pixmap.html#Pixmap.pdfocr_save
+                    # [2]: https://github.com/pymupdf/PyMuPDF/blob/0368e56cfa6afb55bcf6c726e7f51a2a16a5ccba/fitz/fitz.i#L308
+                    sys.modules["fitz.fitz"].TESSDATA_PREFIX = get_tessdata_dir()  # type: ignore [attr-defined]
+
                     page_pdf_bytes = pixmap.pdfocr_tobytes(
                         compress=True,
                         language=ocr_lang,

dev_scripts/dangerzone

@@ -4,10 +4,6 @@
 import os
 import sys
 
-# XXX workaround lack of tessdata path arg for PyMuPDF < v1.22.5
-# for context see https://github.com/freedomofpress/dangerzone/issues/682
-os.environ["TESSDATA_PREFIX"] = os.environ.get("TESSDATA_PREFIX", "/usr/share/tesseract/tessdata")
-
 # Load dangerzone module and resources from the source code tree
 sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
 sys.dangerzone_dev = True