--- In firebird-support@yahoogroups.com, "unordained" <unordained_00@...> wrote:

> I suppose there's the possibility of index corruption. You could test that by
> making sure the optimizer can't use the index:
>
> and (ARTICLE.ISACTIVE || '' = 'F')
>
> or something similar. The only time I've seen an index return incorrect results
> was when I indexed a COMPUTED BY expression that depended on CURRENT_DATE. (A
> terrible idea, by the way, but I was just messing around.) There are index-
> related bugs in JIRA, but they're relatively rare, and it'd be odd for it to be
> there in 2.0, not 2.1, and be back in 2.5...

Currently, there is no index on ISACTIVE. So that cannot be the problem.

>
> For testing purposes, could you add .ISACTIVE to the list of return-columns
> inside your trigger, test its value right after the select, and raise an
> exception if it's 'T' when it shouldn't be?
>

I tried that. No exception occurs, even when it's taking a wrong NEWLOCATION value

> You could also send us the non-simplified version of the trigger, as it could be
> a problem that only appears in the complex version, possibly unrelated to your
> core logic. A typo, even.

I can do that, in a minute or so.

Hans