Subject | Re: Compound indices created under Firebird 2.5.1 |
---|---|
Author | Dmitry Yemanov |
Post date | 2014-03-26T11:54:16Z |
25.03.2014 13:57, stefanschultze560@... wrote:
long time", if you want to be on the safe side.
problems as long you don't have NULLs in that index.
Dmitry
>I'd say that rebuilding a single index is unlikely to take "very very
> From what I see in the corresponding bug tracker cases and version
> control commits, only compound indices containing NULL values seem to be
> affected. We only have one single compound index, and this doesn't
> contain NULL values at all.
long time", if you want to be on the safe side.
> Also, a comment i n Firebird commit 57074 sais the following:Yes, they work the same [incorrect] way. But you may not notice the
>
> Restored the ODS level compatibility with v2.5.1 index keys, while
> using the "old good" index key format in new indices. It allows to
> claim that CORE-3675 and CORE-3853 are fixed (after migration via
> backup/restore) but still operate almost correctly with keys created
> in v2.5.1. This is *much* better than silently returning wrong data
> if a v2.5.1 database is used with v2.5.2 without backup/restore.
>
> Which I understand causes that 2.5.1 indexes work the same way in 2.5.2
> as they worked in 2.5.1.
problems as long you don't have NULLs in that index.
> So, am I correct that in effect nothing really has to be done by ourIt looks so, in your particular case.
> customers when upgrading to 2.5.2?
Dmitry