Subject Re: [firebird-support] Re: firebird deadlock vs isc_tpb_wait/etc. issue, or ?
Author Dmitry Yemanov
learntrade wrote:
>> It reads the record header, then waits until the concurrent transaction
>> commits or rolls back and then re-reads the record (including fetching
>> record data).
> So, it doesn't really read the latest version rec at time of
> (re-)attempt - it (repeatedly) attempts to re-read the rec that was
> the first one it tried?

No, the read version is always the latest one. When I write "record", I
mean the versions chain. Reading the record implies chasing all existing
versions and choosing the latest one.