Re: "BEGIN TRANSACTION" and "START TRANSACTION": different error handling - Mailing list pgsql-hackers

From daveg
Subject Re: "BEGIN TRANSACTION" and "START TRANSACTION": different error handling
Date
Msg-id 20090924120916.GG27227@sonic.net
Whole thread Raw
In response to Re: "BEGIN TRANSACTION" and "START TRANSACTION": different error handling  (Hannu Krosing <hannu@krosing.net>)
List pgsql-hackers
On Thu, Sep 24, 2009 at 12:16:43PM +0300, Hannu Krosing wrote:
> "I expect the transaction is aborted and rollback is executed
> automatically." - this is not how postgreSQL behaves. PostgreSQL needs
> an explicit end of transaction from client, either COMMIT; or ROLLBACK;
> 
> when run from psql, they both act the same, except the string returned
> 
> hannu=# begin transaction;
> BEGIN
> hannu=# select 1/0;
> ERROR:  division by zero
> hannu=# select 1/0;
> ERROR:  current transaction is aborted, commands ignored until end of
> transaction block
> hannu=# abort;
> ROLLBACK
> hannu=# start transaction;
> START TRANSACTION
> hannu=# select 1/0;
> ERROR:  division by zero
> hannu=# select 1/0;
> ERROR:  current transaction is aborted, commands ignored until end of
> transaction block
> hannu=# abort;
> ROLLBACK
> 
> I suspect, that psqlodbc is the one doing the automatic rollback and it
> seems to rely on reply "BEGIN" to establish an in-transaction state.
> 
> so when "start transaction;" returns "START TRANSACTION" instead of
> "BEGIN", psqlodbc does not realise that it is in transaction and does
> not initiate the automatic rollback.

Well. I'd always thought BEGIN and START were syntactic Aspartame and had
the same underlying implementation. So this is a surprise. Why do they
return a different status?

-dg
`
-- 
David Gould       daveg@sonic.net      510 536 1443    510 282 0869
If simplicity worked, the world would be overrun with insects.


pgsql-hackers by date:

Previous
From: Jan Urbański
Date:
Subject: Re: [PATCH] DefaultACLs
Next
From: Petr Jelinek
Date:
Subject: Re: [PATCH] DefaultACLs