Bugzilla

The crash with Dana's file is not platform specific, and even causes oggz-dump to hit the same assertion.  The assert doesn't make any sense to me, like Robert says, there's nothing that ensures packet_buffer is empty between oggz_read_sync and oggz_close other than the regular processing path.  The oggz_dlist_delete call immediately after the assert frees everything that may remain in the list, so the assert doesn't protect an unhandled situation.

It seems like it's possible to hit this with truncated files, otherwise malformed files, or with valid files in shutdown situations (causing the reader to return an error).

Looking at the history, the assert and delete lines were added in the same commit back in October 2007.

Attached patch removes the assert and adds Dana's file as an error test.