[fix/opti] download: clean-up, fix, optimization

An optimization is to avoid temporary file when possible: a rename
(or hard link) is not always possible (different mount points).

Another one is to not check md5sum twice when using cache file.

Fall-back mode is ignored if an MD5 checksum is given.

In case of checksum mismatch for a cached path, remove it and
download again, mainly to cover the following cases:
- the url content changes and the user updates the checksum
- buildout killed while downloading directly to cache
  (see above optimization)
- shutil.copyfile is interrupted

src/zc/buildout/download.py

@@ -54,9 +54,9 @@ import os
 import os.path
 import re
 import shutil
-import sys
 import tempfile
 import zc.buildout
+from .rmtree import rmtree
 
 
 class ChecksumError(zc.buildout.UserError):
@@ -74,7 +74,8 @@ class Download(object):
     cache: path to the download cache (excluding namespaces)
     namespace: namespace directory to use inside the cache
     offline: whether to operate in offline mode
-    fallback: whether to use the cache as a fallback (try downloading first)
+    fallback: whether to use the cache as a fallback (try downloading first),
+              when an MD5 checksum is not given
     hash_name: whether to use a hash of the URL as cache file name
     logger: an optional logger to receive download-related log messages
 
@@ -107,7 +108,8 @@ class Download(object):
         if self.download_cache is not None:
             return os.path.join(self.download_cache, self.namespace or '')
 
-    def __call__(self, url, md5sum=None, path=None):
+    @property
+    def __call__(self):
         """Download a file according to the utility's configuration.
 
         url: URL to download
@@ -117,19 +119,14 @@ class Download(object):
         Returns the path to the downloaded file.
 
         """
-        if self.cache:
-            local_path, is_temp = self.download_cached(url, md5sum)
-        else:
-            local_path, is_temp = self.download(url, md5sum, path)
+        return self.download_cached if self.cache else self.download
 
-        return locate_at(local_path, path), is_temp
-
-    def download_cached(self, url, md5sum=None):
+    def download_cached(self, url, md5sum=None, path=None):
         """Download a file from a URL using the cache.
 
-        This method assumes that the cache has been configured. Optionally, it
-        raises a ChecksumError if a cached copy of a file has an MD5 mismatch,
-        but will not remove the copy in that case.
+        This method assumes that the cache has been configured.
+        If a cached copy of a file has an MD5 mismatch, download
+        and update the cache on success.
 
         """
         if not os.path.exists(self.download_cache):
@@ -146,26 +143,43 @@ class Download(object):
 
         self.logger.debug('Searching cache at %s' % cache_dir)
         if os.path.exists(cached_path):
-            is_temp = False
-            if self.fallback:
-                try:
-                    _, is_temp = self.download(url, md5sum, cached_path)
-                except ChecksumError:
+            if check_md5sum(cached_path, md5sum):
+                if md5sum or not self.fallback:
+                    self.logger.debug('Using cache file %s', cached_path)
+                    return locate_at(cached_path, path), False
+            else:
+                self.logger.warning(
+                    'MD5 checksum mismatch for cached download from %r at %r',
+                    url, cached_path)
+            # Don't download directly to cached_path to minimize
+            # the probability to alter old data if download fails.
+            try:
+                path, is_temp = self.download(url, md5sum, path)
+            except ChecksumError:
+                raise
+            except Exception:
+                if md5sum:
                     raise
-                except Exception:
-                    pass
-
-            if not check_md5sum(cached_path, md5sum):
-                raise ChecksumError(
-                    'MD5 checksum mismatch for cached download '
-                    'from %r at %r' % (url, cached_path))
-            self.logger.debug('Using cache file %s' % cached_path)
+                self.logger.debug("Fallback to cache using %s",
+                                  cached_path, exception=1)
+            else:
+                samefile = getattr(os.path, 'samefile', None)
+                if not (samefile and samefile(path, cached_path)):
+                    # update cache
+                    try:
+                        os.remove(cached_path)
+                    except OSError as e:
+                        if e.errno != errno.EISDIR:
+                            raise
+                        rmtree(cached_path)
+                    locate_at(path, cached_path)
+                return path, is_temp
         else:
             self.logger.debug('Cache miss; will cache %s as %s' %
                               (url, cached_path))
-            _, is_temp = self.download(url, md5sum, cached_path)
+            self.download(url, md5sum, cached_path)
 
-        return cached_path, is_temp
+        return locate_at(cached_path, path), False
 
     def download(self, url, md5sum=None, path=None):
         """Download a file from a URL to a given or temporary path.
@@ -196,27 +210,25 @@ class Download(object):
                 "Couldn't download %r in offline mode." % url)
 
         self.logger.info('Downloading %s' % url)
-        handle, tmp_path = tempfile.mkstemp(prefix='buildout-')
-        os.close(handle)
+        tmp_path = path
+        cleanup = True
         try:
+            if not path:
+                handle, tmp_path = tempfile.mkstemp(prefix='buildout-')
+                os.close(handle)
             tmp_path, headers = urlretrieve(url, tmp_path)
             if not check_md5sum(tmp_path, md5sum):
                 raise ChecksumError(
                     'MD5 checksum mismatch downloading %r' % url)
-        except IOError:
-            e = sys.exc_info()[1]
-            os.remove(tmp_path)
-            raise zc.buildout.UserError("Error downloading extends for URL "
-                              "%s: %s" % (url, e))
-        except Exception:
-            os.remove(tmp_path)
-            raise
-
-        if path:
-            shutil.move(tmp_path, path)
-            return path, False
-        else:
-            return tmp_path, True
+            cleanup = False
+        except IOError as e:
+            raise zc.buildout.UserError("Error downloading %s: %s"
+                                        % (url, e))
+        finally:
+            if cleanup and tmp_path:
+                remove(tmp_path)
+
+        return tmp_path, not path
 
     def filename(self, url):
         """Determine a file name from a URL according to the configuration.

src/zc/buildout/tests/download.txt

@@ -165,14 +165,6 @@ the file on the server to see this:
 >>> cat(path)
 This is a foo text.
 
-If we specify an MD5 checksum for a file that is already in the cache, the
-cached copy's checksum will be verified:
-
->>> download(server_url+'foo.txt', md5('The wrong text.'.encode()).hexdigest())
-Traceback (most recent call last):
-ChecksumError: MD5 checksum mismatch for cached download
-               from 'http://localhost/foo.txt' at '/download-cache/foo.txt'
-
 Trying to access another file at a different URL which has the same base name
 will result in the cached copy being used:
 
@@ -184,6 +176,14 @@ will result in the cached copy being used:
 >>> cat(path)
 This is a foo text.
 
+If we specify an MD5 checksum for a file that is already in the cache, the
+cached copy's checksum will be verified and the cache will be refreshed:
+
+>>> path, is_temp = download(server_url+'foo.txt', md5('The wrong text.'.encode()).hexdigest())
+>>> is_temp
+True
+>>> remove(path)
+
 Given a target path for the download, the utility will provide a copy of the
 file at that location both when first downloading the file and when using a
 cached copy:
@@ -259,7 +259,7 @@ If the file is completely missing it should notify the user of the error:
 >>> download(server_url+'bar.txt') # doctest: +NORMALIZE_WHITESPACE +ELLIPSIS
 Traceback (most recent call last):
 ...
-UserError: Error downloading extends for URL http://localhost/bar.txt:
+UserError: Error downloading http://localhost/bar.txt:
 ...404...
 >>> ls(cache)
 
@@ -442,18 +442,22 @@ However, when downloading the file normally with the cache being used in
 fall-back mode, the file will be downloaded from the net and the cached copy
 will be replaced with the new content:
 
->>> cat(download(server_url+'foo.txt')[0])
+>>> path, is_temp = download(server_url+'foo.txt')
+>>> cat(path)
 The wrong text.
 >>> cat(cache, 'foo.txt')
 The wrong text.
+>>> is_temp
+True
+>>> remove(path)
 
-When trying to download a resource whose checksum does not match, the cached
-copy will neither be used nor overwritten:
+Fall-back mode is meaningless if md5sum is given. If the checksum of the
+cached copy matches, the resource is not downloaded:
 
 >>> write(server_data, 'foo.txt', 'This is a foo text.')
->>> download(server_url+'foo.txt', md5('The wrong text.'.encode()).hexdigest())
-Traceback (most recent call last):
-ChecksumError: MD5 checksum mismatch downloading 'http://localhost/foo.txt'
+>>> path, is_temp = download(server_url+'foo.txt', md5('The wrong text.'.encode()).hexdigest())
+>>> print_(path)
+/download-cache/foo.txt
 >>> cat(cache, 'foo.txt')
 The wrong text.