Fujii Masao wrote:
> This assertion failure derives from the recent refactoring of ReadRecord().
> Which changed the startup process so as to re-fetch the last applied WAL
> record at the end of recovery from the buffer instead of the WAL file if it's
> stored in the buffer. In this case, the last applied WAL file remains closed.
> So readFile (which ought to have been the file descriptor of that WAL file)
> might be -1 in exitArchiveRecovery().
>
> In the now, that assertion is obsolete. So I attached the patch that removes
> the assert() from exitArchiveRecovery().
Committed.
-- Heikki Linnakangas EnterpriseDB http://www.enterprisedb.com