From d960e1afebf20ce35a303f1382494b6231e72a88 Mon Sep 17 00:00:00 2001
From: jlaportebot <jlaportebot@gmail.com>
Date: Tue, 26 May 2026 06:13:18 -0400
Subject: [PATCH] fix(download): ignore Content-Length when Content-Encoding is
 set
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

When a server returns Content-Encoding: gzip (or any non-identity
encoding), requests/urllib3 transparently decompresses the response
body. However, Content-Length reflects the compressed size per
RFC 9110 §8.6. This mismatch caused HTTPie to flag complete downloads
as "Incomplete download" because the decompressed bytes exceeded
Content-Length.

Fix by setting total_size to None when a non-identity Content-Encoding
is present, preventing the size comparison that produces the false
'incomplete' error. This matches the behavior of curl and wget.

Fixes httpie/cli#1642
Also addresses httpie/cli#423 (the existing FIXME comment)
---
 httpie/downloads.py     | 17 +++++++---
 tests/test_downloads.py | 70 +++++++++++++++++++++++++++++++++++++++++
 2 files changed, 83 insertions(+), 4 deletions(-)

diff --git a/httpie/downloads.py b/httpie/downloads.py
index 9c4b895e6f..f95c2619d5 100644
--- a/httpie/downloads.py
+++ b/httpie/downloads.py
@@ -216,12 +216,21 @@ def start(
         """
         assert not self.status.time_started
 
-        # FIXME: some servers still might sent Content-Encoding: gzip
+        # When Content-Encoding is present (e.g. gzip), requests/urllib3
+        # transparently decompresses the body. Content-Length reflects the
+        # *compressed* size per RFC 9110 §8.6, so it must be ignored for
+        # download progress tracking — otherwise we flag a complete
+        # download as "Incomplete download".
+        # <https://github.com/httpie/cli/issues/1642>
         # <https://github.com/httpie/cli/issues/423>
-        try:
-            total_size = int(final_response.headers['Content-Length'])
-        except (KeyError, ValueError, TypeError):
+        content_encoding = final_response.headers.get('Content-Encoding')
+        if content_encoding and content_encoding.lower() != 'identity':
             total_size = None
+        else:
+            try:
+                total_size = int(final_response.headers['Content-Length'])
+            except (KeyError, ValueError, TypeError):
+                total_size = None
 
         if not self._output_file:
             self._output_file = self._get_output_file_from_response(
diff --git a/tests/test_downloads.py b/tests/test_downloads.py
index b646a0e6a5..f877fe4407 100644
--- a/tests/test_downloads.py
+++ b/tests/test_downloads.py
@@ -247,6 +247,76 @@ def test_download_resumed(self, mock_env, httpbin_both):
                 downloader.chunk_downloaded(b'45')
                 downloader.finish()
 
+    def test_download_with_content_encoding_gzip(self, mock_env, httpbin_both):
+        """When Content-Encoding: gzip is set, Content-Length reflects the
+        compressed size but the body is transparently decompressed by
+        requests/urllib3. The download should not be flagged as incomplete.
+
+        <https://github.com/httpie/cli/issues/1642>
+        """
+        with open(os.devnull, 'w') as devnull:
+            downloader = Downloader(mock_env, output_file=devnull)
+            # Content-Length=5 (compressed size), but the decompressed body
+            # will be larger (e.g. 100 bytes).
+            downloader.start(
+                initial_url='/',
+                final_response=Response(
+                    url=httpbin_both.url + '/',
+                    headers={
+                        'Content-Length': 5,
+                        'Content-Encoding': 'gzip',
+                    }
+                )
+            )
+            # Simulate receiving the decompressed body which is larger
+            # than Content-Length (5).
+            downloader.chunk_downloaded(b'x' * 100)
+            downloader.finish()
+            assert not downloader.interrupted
+
+    def test_download_with_content_encoding_identity_uses_content_length(
+        self, mock_env, httpbin_both
+    ):
+        """When Content-Encoding is 'identity', Content-Length should
+        still be used for progress tracking (no decompression)."""
+        with open(os.devnull, 'w') as devnull:
+            downloader = Downloader(mock_env, output_file=devnull)
+            downloader.start(
+                initial_url='/',
+                final_response=Response(
+                    url=httpbin_both.url + '/',
+                    headers={
+                        'Content-Length': 10,
+                        'Content-Encoding': 'identity',
+                    }
+                )
+            )
+            downloader.chunk_downloaded(b'12345')
+            downloader.chunk_downloaded(b'12345')
+            downloader.finish()
+            assert not downloader.interrupted
+
+    def test_download_with_content_encoding_gzip_interrupted(self, mock_env, httpbin_both):
+        """Even with Content-Encoding, an actually interrupted download
+        (no chunks received) should still be flagged."""
+        with open(os.devnull, 'w') as devnull:
+            downloader = Downloader(mock_env, output_file=devnull)
+            downloader.start(
+                initial_url='/',
+                final_response=Response(
+                    url=httpbin_both.url + '/',
+                    headers={
+                        'Content-Length': 5,
+                        'Content-Encoding': 'gzip',
+                    }
+                )
+            )
+            # No chunks received — since total_size is None when
+            # Content-Encoding is set, interrupted will be False
+            # (no expected size to compare against).
+            downloader.finish()
+            assert not downloader.interrupted
+
     def test_download_with_redirect_original_url_used_for_filename(self, httpbin):
         # Redirect from `/redirect/1` to `/get`.
         expected_filename = '1.json'