Tiago,

Unfortunately I think your test script is ambiguous as it does not say when
to commit changes.

I ran through your tests as follows and unless I commit (1) after the select
and before the update (i.e. after (2) has added the column and committed) I
get the error:
Dynamic SQL Error.
SQL error code = -206.
Column unknown.
FIELD2.
At line 1, column 19.

Can you give me the test again with:
1) The transaction isolation level
2) The commits e.g. (1) commit
3) The start transaction (in most tools the point of start is the first
command after a commit).

What tool are you using?

Cheers,

JAC

"Tiago Costa" <tiago@...> wrote in message
news:3E4BF5F1.3050608@......

> Jason,
>
> I tried that and it worked ok. Thanks.
>
> But I wondered what would happen if I start a transaction in a
> connection, do some updates/inserts, then in another connection, make
> the metadata changes and in the first connection (in the same
> transaction) use the new fields that connection 2 has created, and the
> rollback...
> What I found out: (test-case)
>
>
> CREATE TABLE table1 (id integer, field1 varchar(1));
> insert into table1 (id,field1) values (1,'')
>
>
> Two different connections to the same database (1) and (2)
>
> (1st connection) Start transaction
>
> (1) SELECT * FROM table1
> id | field1
> -------------
> 1 | ""
>
> (1) UPDATE table1 SET field1='T' where id=1
>
> (2nd connection) ALTER TABLE table1 add field2 VARCHAR(10)
>
> (1) UPDATE table1 SET field2="test" where id=1
>
> (1) SELECT * FROM table1
> id | field1 | field2 |
> ------------------------
> 1 | "T" | "test" |
>
>
> (1) ROLLBACK
>
> (1) SELECT * FROM table1
>
> id | field1 | field2 |
> ------------------------
> 1 | "T" | "" |
>
> the "field1" should also be blank...
>
> I know that's **VERY** unprobable to happen, but...
>
> thanks again,
>
> Tiago
>
>
> Jason Chapman (JAC2) wrote:
>
> >No it doesn't cause corruption AFAIK - I do this regularly.
> >
> >When we had > 100 concurrent users I used to wait until out of hours,

just

> >in case I goosed it and had to regress.
> >
> >JAC.
> >
> >"Tiago Costa" <tiago@...> wrote in message
> >news:3E4BD1CD.1080103@......
> >
> >
> >>Hello,
> >>
> >> I wonder if it is safe to do several "ALTER TABLE xxx ADD fname type",
> >>"DROP INDEX", "CREATE INDEX" in a live database or it can cause
> >>
> >>
> >corruption?
> >
> >
> >>TIA,
> >>Tiago
> >>
> >>
> >>
> >>
> >>To unsubscribe from this group, send an email to:
> >>ib-support-unsubscribe@egroups.com
> >>
> >>
> >>
> >>Your use of Yahoo! Groups is subject to

http://docs.yahoo.com/info/terms/

> >>
> >>
> >>
> >>
> >>
> >
> >
> >
> >
> >To unsubscribe from this group, send an email to:
> >ib-support-unsubscribe@egroups.com
> >
> >
> >
> >Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/
> >
> >
> >
> >
> >
> >
>
> --
> Tiago Silva Costa
>
> ECALC Sistemas de Computacao Ltda
> Tel:(11)3898-1602 Fax:(11)3063-0359
> email: tiago@...
> http://www.ecalc.com.br
>
>
>
>
>
>
>
> [Non-text portions of this message have been removed]
>
>
>
> To unsubscribe from this group, send an email to:
> ib-support-unsubscribe@egroups.com
>
>
>
> Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/
>
>
>