Detect removed files and reflect the state in chrome://downloads and the download shelf

chrome/browser/download/download_util.cc

Index: chrome/browser/download/download_util.cc
diff --git a/chrome/browser/download/download_util.cc b/chrome/browser/download/download_util.cc
index 05751ded1ff230e8b028cb2f9f6e1809025a5dfa..98ecfad21f94a32d31099b4704f71b60e7f7b7bf 100644
--- a/chrome/browser/download/download_util.cc
+++ b/chrome/browser/download/download_util.cc
@@ -664,7 +664,19 @@ DictionaryValue* CreateDownloadItemValue(DownloadItem* download, int id) {
     if (download->safety_state() == DownloadItem::DANGEROUS) {
       file_value->SetString("state", "DANGEROUS");
     } else {
-      file_value->SetString("state", "COMPLETE");
+      bool path_exists = true;
+#if defined(OS_WIN)
+      if (GetFileAttributes(
+              download->GetTargetFilePath().value().c_str()) ==
+          INVALID_FILE_ATTRIBUTES)
+        path_exists = false;
+#elif defined(OS_POSIX)
+      if (access(download->GetTargetFilePath().value().c_str(), F_OK) != 0)
+        path_exists = false;
+#else
+      NOTIMPLEMENTED();
+#endif
+      file_value->SetString("state", path_exists ? "COMPLETE" : "REMOVED");

[Randy Smith (Not in Mondays), 2011/04/27 18:21:21]

All file system accesses need to happen on the FILE thread. 
CreateDownloadItemValue isn't restricted to the file thread, and I believe (from
a cursory scan of the call sites) is usually called on the UI thread.  So this
won't work.

[Paweł Hajdan Jr., 2011/04/29 08:43:41]

Exactly. And using SetAllow... or re-implementing the code is really just
bypassing checks intended to catch blocking IO on wrong threads.

Please note we also cannot jump through the FILE thread in a way that would make
the UI thread block or wait on the FILE thread.

I think the check for file existence should be done asynchronously.

     }
   }