Cached or already-downloaded files with bad hashes are ignored.

pip/download.py

@@ -573,7 +573,25 @@ def unpack_http_url(link, location, download_cache, download_dir=None):
         if link.hash:
             download_hash = _get_hash_from_file(temp_location, link)
         logger.notify('File was already downloaded %s' % already_downloaded)
-    else:
+
+    if download_hash:
+        try:
+            _check_hash(download_hash, link)
+        except HashMismatch:
+            logger.warn(
+                'Cached or previously-downloaded file %s has bad hash, '
+                're-downloading.' % temp_location
+                )
+            temp_location = None
+            if already_downloaded:
+                os.unlink(already_downloaded)
+            already_downloaded = None
+            if already_cached:
+                os.unlink(cache_file)
+                os.unlink(cache_content_type_file)
+                already_cached = False
+
+    if not temp_location:
         resp = _get_response_from_url(target_url, link)
         content_type = resp.info().get('content-type', '')
         filename = link.filename  # fallback
@@ -595,6 +613,7 @@ def unpack_http_url(link, location, download_cache, download_dir=None):
                 filename += ext
         temp_location = os.path.join(temp_dir, filename)
         download_hash = _download_url(resp, link, temp_location)
+
     if link.hash and link.hash_name:
         _check_hash(download_hash, link)
     if download_dir and not already_downloaded:
@@ -602,7 +621,7 @@ def unpack_http_url(link, location, download_cache, download_dir=None):
     unpack_file(temp_location, location, content_type, link)
     if cache_file and cache_file != temp_location:
         cache_download(cache_file, temp_location, content_type)
-    if cache_file is None and not already_downloaded:
+    if not (already_cached or already_downloaded):
         os.unlink(temp_location)
     os.rmdir(temp_dir)
 

pip/util.py

@@ -582,7 +582,6 @@ def cache_download(target_file, temp_location, content_type):
     fp = open(target_file+'.content-type', 'w')
     fp.write(content_type)
     fp.close()
-    os.unlink(temp_location)
 
 
 def unpack_file(filename, location, content_type, link):

tests/unit/test_download.py

@@ -1,9 +1,11 @@
+import hashlib
 import os
 from shutil import rmtree
 from tempfile import mkdtemp
 
-import pip
 from mock import patch
+import pip
+from pip.backwardcompat import urllib, BytesIO, b
 from pip.download import (_get_response_from_url as _get_response_from_url_original,
                           path_to_url2, unpack_http_url, URLOpener)
 from pip.index import Link
@@ -36,13 +38,78 @@ def test_user_agent():
     assert user_agent[1].startswith("pip/%s" % pip.__version__)
 
 
-def test_bad_cache_checksum():
+def _write_file(fn, contents):
+    with open(fn, 'w') as fh:
+        fh.write(contents)
+
+
+class MockResponse(object):
+    def __init__(self, contents):
+        self._io = BytesIO(contents)
+
+    def read(self, *a, **kw):
+        return self._io.read(*a, **kw)
+
+
+@patch('pip.download.unpack_file')
+@patch('pip.download._get_response_from_url')
+def test_unpack_http_url_bad_cache_checksum(mock_get_response, mock_unpack_file):
     """
     If cached download has bad checksum, re-download.
     """
+    base_url = 'http://www.example.com/somepackage.tgz'
+    contents = b('downloaded')
+    download_hash = hashlib.new('sha1', contents)
+    link = Link(base_url + '#sha1=' + download_hash.hexdigest())
+    response = mock_get_response.return_value = MockResponse(contents)
+    response.info = lambda: {'content-type': 'application/x-tar'}
+    response.geturl = lambda: base_url
+
+    cache_dir = mkdtemp()
+    try:
+        cache_file = os.path.join(cache_dir, urllib.quote(base_url, ''))
+        cache_ct_file = cache_file + '.content-type'
+        _write_file(cache_file, 'some contents')
+        _write_file(cache_ct_file, 'application/x-tar')
+
+        unpack_http_url(link, 'location', download_cache=cache_dir)
+
+        # despite existence of cached file with bad hash, downloaded again
+        mock_get_response.assert_called_once_with(base_url, link)
+        # cached file is replaced with newly downloaded file
+        with open(cache_file) as fh:
+            assert fh.read() == 'downloaded'
+
+    finally:
+        rmtree(cache_dir)
 
 
-def test_bad_already_downloaded_checksum():
+@patch('pip.download.unpack_file')
+@patch('pip.download._get_response_from_url')
+def test_unpack_http_url_bad_downloaded_checksum(mock_get_response, mock_unpack_file):
     """
     If already-downloaded file has bad checksum, re-download.
     """
+    base_url = 'http://www.example.com/somepackage.tgz'
+    contents = b('downloaded')
+    download_hash = hashlib.new('sha1', contents)
+    link = Link(base_url + '#sha1=' + download_hash.hexdigest())
+    response = mock_get_response.return_value = MockResponse(contents)
+    response.info = lambda: {'content-type': 'application/x-tar'}
+    response.geturl = lambda: base_url
+
+    download_dir = mkdtemp()
+    try:
+        downloaded_file = os.path.join(download_dir, 'somepackage.tgz')
+        _write_file(downloaded_file, 'some contents')
+
+        unpack_http_url(link, 'location', download_cache=None, download_dir=download_dir)
+
+        # despite existence of downloaded file with bad hash, downloaded again
+        mock_get_response.assert_called_once_with(base_url, link)
+        # cached file is replaced with newly downloaded file
+        with open(downloaded_file) as fh:
+            assert fh.read() == 'downloaded'
+
+    finally:
+        rmtree(download_dir)