Don't consider restart when the test action already finished
authorlurchi <lurchi@strangeplace.net>
Mon, 4 Sep 2017 07:17:45 +0000 (09:17 +0200)
committerlurchi <lurchi@strangeplace.net>
Mon, 4 Sep 2017 07:17:45 +0000 (09:17 +0200)
In the old scheduler implementation tasks with different priorities were
run in the same event loop iteration, in the new implementation this
cannot be guaranteed anymore, as it's driver-dependent.
For the FS tests this means that when adding both an 'abort' task and a
'restart' task (the latter with higher priority) the 'restart' task
didn't take effect with the old implementation but does with the new
implementation (and leads to an assertion failure).
Not scheduling the 'restart' task when the action (download/publish) has
already finished successfully solves this issue.

src/fs/test_fs_download_persistence.c
src/fs/test_fs_publish_persistence.c

index 76a1ea91104605963c8091460b6b77935639cee3..8f27e82af579d81989cfb95a1239e3e7dc245642 100644 (file)
@@ -179,7 +179,6 @@ progress_cb (void *cls, const struct GNUNET_FS_ProgressInfo *event)
                               GNUNET_FS_DOWNLOAD_OPTION_NONE, "download", NULL);
     break;
   case GNUNET_FS_STATUS_DOWNLOAD_COMPLETED:
-    consider_restart (event->status);
     printf ("Download complete,  %llu kbps.\n",
             (unsigned long long) (FILESIZE * 1000000LL /
                                   (1 +
index be9006d4233f80d8301ef7b64bb2ba729335bbf2..103ca01b8bae34cf509dab4f69cd901dda547e03 100644 (file)
@@ -134,7 +134,6 @@ progress_cb (void *cls,
   switch (event->status)
   {
   case GNUNET_FS_STATUS_PUBLISH_COMPLETED:
-    consider_restart (event->status);
     ret = event->value.publish.cctx;
     printf ("Publish complete,  %llu kbps.\n",
             (unsigned long long) (FILESIZE * 1000000LL /