Make PyMuPDF always log to stderr

dangerzone/conversion/doc_to_pixels.py

@@ -3,6 +3,17 @@
 import sys
 from typing import Dict, Optional
 
+# XXX: PyMUPDF logs to stdout by default [1]. The PyMuPDF devs provide a way [2] to log to
+# stderr, but it's based on environment variables. These envvars are consulted at import
+# time [3], so we have to set them here, before we import `fitz`.
+#
+# [1] https://github.com/freedomofpress/dangerzone/issues/877
+# [2] https://github.com/pymupdf/PyMuPDF/issues/3135#issuecomment-1992625724
+# [3] https://github.com/pymupdf/PyMuPDF/blob/9717935eeb2d50d15440d62575878214226795f9/src/__init__.py#L62-L63
+os.environ["PYMUPDF_MESSAGE"] = "fd:2"
+os.environ["PYMUPDF_LOG"] = "fd:2"
+
+
 import fitz
 import magic
 

dangerzone/conversion/pixels_to_pdf.py

@@ -16,6 +16,16 @@
 
 from .common import DEFAULT_DPI, DangerzoneConverter, get_tessdata_dir, running_on_qubes
 
+# XXX: PyMUPDF logs to stdout by default [1]. The PyMuPDF devs provide a way [2] to log to
+# stderr, but it's based on environment variables. These envvars are consulted at import
+# time [3], so we have to set them here, before we import `fitz`.
+#
+# [1] https://github.com/freedomofpress/dangerzone/issues/877
+# [2] https://github.com/pymupdf/PyMuPDF/issues/3135#issuecomment-1992625724
+# [3] https://github.com/pymupdf/PyMuPDF/blob/9717935eeb2d50d15440d62575878214226795f9/src/__init__.py#L62-L63
+os.environ["PYMUPDF_MESSAGE"] = "fd:2"
+os.environ["PYMUPDF_LOG"] = "fd:2"
+
 
 class PixelsToPDF(DangerzoneConverter):
     async def convert(
@@ -50,14 +60,13 @@ async def convert(
             # The first few operations happen on a per-page basis.
             page_size = len(untrusted_rgb_data)
             total_size += page_size
-            with contextlib.redirect_stdout(io.StringIO()):
-                pixmap = fitz.Pixmap(
-                    fitz.Colorspace(fitz.CS_RGB),
-                    width,
-                    height,
-                    untrusted_rgb_data,
-                    False,
-                )
+            pixmap = fitz.Pixmap(
+                fitz.Colorspace(fitz.CS_RGB),
+                width,
+                height,
+                untrusted_rgb_data,
+                False,
+            )
             pixmap.set_dpi(DEFAULT_DPI, DEFAULT_DPI)
             if ocr_lang:  # OCR the document
                 self.update_progress(

tests/test_large_set.py

@@ -75,7 +75,12 @@ def run_doc_test(self, doc: Path, tmp_path: Path) -> None:
             stdout=subprocess.PIPE,
             stderr=subprocess.STDOUT,
         )
-        out, _ = p.communicate()
+        try:
+            # Set a global timeout of 5 minutes for the processing of a document.
+            # This is hacky way to sidestep https://github.com/freedomofpress/dangerzone/issues/878
+            out, _ = p.communicate(timeout=5 * 60)
+        except subprocess.TimeoutExpired:
+            print(f"*** TIMEOUT EXCEEDED FOR DOCUMENT '{doc}' ***")
         from strip_ansi import strip_ansi
 
         print(strip_ansi(out.decode()))