Treat ERR_CONNECTION_CLOSED as end-of-data marker for downloads.

chrome/browser/download/download_manager.cc

Index: chrome/browser/download/download_manager.cc
diff --git a/chrome/browser/download/download_manager.cc b/chrome/browser/download/download_manager.cc
index ccedecf5b5fdb06bcc85c8de983d92fdd149367f..2c28f5cdcc4355de74d7dde35281c821dfd11310 100644
--- a/chrome/browser/download/download_manager.cc
+++ b/chrome/browser/download/download_manager.cc
@@ -539,7 +539,12 @@ void DownloadManager::OnResponseCompleted(int32 download_id,
                                           int os_error,
                                           const std::string& hash) {
   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
-  if (os_error == 0) {
+  // Some web sites lie about the size of the data (for example, using gzip
+  // encoding and specifying the size of the expanded data) and disconnect
+  // when the data is done.  Unfortunately, this means that if we get a
+  // disconnect and the data hasn't been completely downloaded we can't tell.
+  // See http://code.google.com/p/chromium/issues/detail?id=78769.
+  if ((os_error == 0) || (os_error == net::ERR_CONNECTION_CLOSED)) {

[cbentzel, 2011/04/21 17:16:41]

We discussed doing this at a lower-layer [in HttpStreamRequest] only in the case
when the response has Connection: close, and not returning ERR_CONNECTION_CLOSED
in that case even if bytes read in body did not equal Content-Length. There's
likely other cases which will result in ERR_CONNECTION_CLOSED here [i.e. it's a
Keep-Alive connection but there's a TCP RST].

You should take a look at the Firefox code to see how they handle that case as
well.

     OnAllDataSaved(download_id, size, hash);
   } else {
     OnDownloadError(download_id, size, os_error);