[06:00:22] --- jaltman/FrogsLeap has left: Disconnected [06:14:22] --- jaltman/FrogsLeap has become available [06:55:22] --- kaduk@mit.edu/barnowl has become available [07:28:07] --- lars.malinowsky has left [07:28:13] --- lars.malinowsky has become available [07:38:38] --- jaltman/FrogsLeap has left: Disconnected [07:41:14] --- mfelliott has left [07:58:21] --- jaltman/FrogsLeap has become available [08:03:48] --- jaltman/FrogsLeap has left: Replaced by new connection [08:03:49] --- jaltman/FrogsLeap has become available [08:28:46] --- jaltman/FrogsLeap has left: Disconnected [08:29:01] --- jaltman/FrogsLeap has become available [08:34:40] --- jaltman/FrogsLeap has left: Disconnected [10:00:27] --- Russ has become available [11:59:29] --- sxw has become available [12:09:17] --- sxw has left [12:27:25] --- jaltman/FrogsLeap has become available [12:31:12] --- Russ has left: Disconnected [12:51:12] --- rra has become available [14:07:48] 1.6.0pre5 appears to have some sort of cache corruption bug with large files. The tail end of a large file is just all nuls. [14:07:51] fs flush doesn't fix it. [14:08:11] Nor does stopping the client, deleting the cache, and starting the client again. When the file is read from the server again, the tail end of the file is all nuls. [14:13:17] The nuls start exactly one character after my chunk size. [14:13:20] I filed a bug. [14:26:14] --- rra has left: Disconnected [14:32:25] --- rra has become available [14:33:30] Reverting 2571b6285d5da8ef62ab38c3a938258ddd7bac4e "Linux: Don't read pages beyond the cache eof" fixes the problem. [17:43:56] --- mdionne has become available [17:51:02] rra: different patch sent through the ticket. what we need to check against the cache file size is the offset into the current chunk [17:51:43] Giving it a try now. [17:54:41] Yup, also seems to work fine. [17:56:13] ok, and this one doesn't bring back the issue I had with tmpfs that the original commit tried to fix [17:59:32] will submit to gerrit shortly [19:28:24] --- mdionne has left [22:08:57] hrm, corruption not good. good thing I keep putting off deploying 1.6.0 client =/ [22:21:38] --- andersk has become available