Read report/label templates in binary mode (#12432)

* Read report/label templates in binary mode

file_from_template read the template in text mode and wrapped the string
in a ContentFile. The ContentFile size then reflected the character count,
which differs from the UTF-8 byte length for any template that contains
multi-byte characters. Strict S3 implementations such as Backblaze B2
validate the Content-Length header and reject the upload with an
IncompleteBody error. AWS S3 tolerates the mismatch, so the problem is
not visible there.

inventree_transfer_order_report.html is 1458 characters but 1460 bytes.
Its upload fails during create_default_reports, the transaction rolls
back, and the default template is recreated and fails again on every
startup.

Use Path.read_bytes() so the ContentFile size matches the uploaded bytes
and the file handle is closed. Add a regression test asserting the size
equals the file's byte length.

* Test file_from_template across text encodings

src/backend/InvenTree/report/apps.py

@@ -96,7 +96,7 @@ class ReportConfig(AppConfig):
             raise FileNotFoundError(
                 'Template file %s does not exist in %s', file_name, file
             )
-        return ContentFile(file.open('r').read(), os.path.basename(file_name))
+        return ContentFile(file.read_bytes(), os.path.basename(file_name))
 
     def create_default_labels(self):
         """Create default label templates."""

src/backend/InvenTree/report/tests.py

@@ -1,7 +1,9 @@
 """Unit testing for the various report models."""
 
 import os
+import tempfile
 from io import StringIO
+from pathlib import Path
 from unittest.mock import patch
 
 from django.apps import apps
@@ -18,6 +20,7 @@ import report.models as report_models
 from build.models import Build
 from common.models import Attachment
 from common.settings import set_global_setting
+from InvenTree.config import get_base_dir
 from InvenTree.unit_test import AdminTestCase, InvenTreeAPITestCase
 from order.models import PurchaseOrder, ReturnOrder, SalesOrder
 from part.models import Part
@@ -1073,3 +1076,66 @@ class URLFetcherTest(TestCase):
         with patch('weasyprint.urls.URLFetcher.fetch', return_value={}):
             self.fetcher.fetch('data:image/png;base64,abc123')
             self.fetcher.fetch('data:text/css;base64,abc123')
+
+
+class DefaultTemplateFileTest(TestCase):
+    """Unit tests for building the default report and label template files."""
+
+    def test_file_size_matches_bytes(self):
+        """file_from_template must size the ContentFile by byte length.
+
+        Reading the template in text mode makes the size a character count,
+        which is smaller than the byte length for multi-byte content and breaks
+        uploads to S3 backends that enforce Content-Length.
+        """
+        config = apps.get_app_config('report')
+        filename = 'inventree_transfer_order_report.html'
+        content_file = config.file_from_template('report', filename)
+        path = get_base_dir().joinpath('report', 'templates', 'report', filename)
+        self.assertEqual(content_file.size, path.stat().st_size)
+
+    def test_file_from_template_is_encoding_agnostic(self):
+        """file_from_template must preserve the exact bytes for any encoding.
+
+        A text-mode read sizes the ContentFile by character count and needs the
+        correct decoder, so non-English content is both mis-sized and at risk of
+        corruption. Reading the raw bytes keeps the file identical to what is on
+        disk, so the size always matches the Content-Length used for S3 uploads.
+        The samples below cover non-ASCII text in a few encodings, including
+        multi-byte content where a character count would not equal the byte
+        length.
+        """
+        config = apps.get_app_config('report')
+
+        # (encoding, text, whether the encoding uses multi-byte characters).
+        # The utf cases hold characters that span more than one byte, so a
+        # character count would not equal the byte length; latin-1 stays
+        # single-byte and is included to show the read is not utf-8 specific.
+        cases = [
+            ('utf-8', 'Système de café ≤ 你好 в наявності 📦', True),
+            ('utf-16', 'Système de café ≤ 你好 в наявності 📦', True),
+            ('latin-1', 'Système de café àéîõü', False),
+        ]
+
+        for encoding, text, multibyte in cases:
+            with self.subTest(encoding=encoding):
+                body = f'<html><body>{text}</body></html>'
+                raw = body.encode(encoding)
+
+                if multibyte:
+                    # A character count would differ from the byte length here
+                    self.assertNotEqual(len(raw), len(body))
+
+                with tempfile.TemporaryDirectory() as tmp:
+                    base = Path(tmp)
+                    template_dir = base.joinpath('report', 'templates', 'report')
+                    template_dir.mkdir(parents=True)
+                    filename = f'encoding_{encoding}.html'
+                    template_dir.joinpath(filename).write_bytes(raw)
+
+                    with patch('report.apps.get_base_dir', return_value=base):
+                        content_file = config.file_from_template('report', filename)
+
+                # Size and content must match the raw bytes exactly
+                self.assertEqual(content_file.size, len(raw))
+                self.assertEqual(content_file.read(), raw)