Hello,
On Wed, Jul 09, 2008 at 11:24:44PM +0200, Guilhem Bichot wrote:
> Hello,
>
> this statement shouldn't use versioning (as it may delete rows). I
> haven't checked if it does, but looking at code... do we have a bug?
I dug a bit, after adding
DBUG_ASSERT(file->lock.type != TL_WRITE_CONCURRENT_INSERT);
in ha_maria::update_row and delete_row:
- LOAD DATA use TL_WRITE_LOW_PRIORITY or TL_WRITE by default, so this
is ok
- but LOAD DATA CONCURRENT uses TL_WRITE_CONCURRENT_INSERT so if I add
a REPLACE, assertion fires in ha_maria::update_row
- INSERT VALUES ON DUPLICATE KEY UPDATE uses TL_WRITE_LOW_PRIORITY or
TL_WRITE, ok
- INSERT SELECT ON DUPLICATE KEY UPDATE uses
TL_WRITE_CONCURRENT_INSERT, assertion fires in ha_maria::update_row.
I'll commit a fix for those.
--
__ ___ ___ ____ __
/ |/ /_ __/ __/ __ \/ / Mr. Guilhem Bichot <guilhem@stripped>
/ /|_/ / // /\ \/ /_/ / /__ MySQL France, Lead Software Engineer
/_/ /_/\_, /___/\___\_\___/ Bordeaux, France
<___/ www.mysql.com
