Thread: BUG #8230: Time interval size overflow
The following bug has been logged on the website: Bug reference: 8230 Logged by: Rok Kralj Email address: rok.kralj@gmail.com PostgreSQL version: 9.2.4 Operating system: Arch Linux Description: = As far as I know, the INTERVAL type is represented using three int32, one for months, days and milliseconds. However, overflow detections seems to work correctly only on these three types; if they get overflown indirectly, they do not get detected: SELECT INTERVAL '3000000000 days' ERROR: interval field value out of range: "3000000000 days" SELECT INTERVAL '2000000000 years' -147483648 years I'm surprised this hasn't been reported yet. The expected behaviour is to raise an error.
On Sun, Jun 16, 2013 at 11:03:53AM +0000, rok.kralj@gmail.com wrote:
> The following bug has been logged on the website:
>
> Bug reference: 8230
> Logged by: Rok Kralj
> Email address: rok.kralj@gmail.com
> PostgreSQL version: 9.2.4
> Operating system: Arch Linux
> Description:
>
> As far as I know, the INTERVAL type is represented using three int32, one
> for months, days and milliseconds.
>
> However, overflow detections seems to work correctly only on these three
> types; if they get overflown indirectly, they do not get detected:
>
> SELECT INTERVAL '3000000000 days'
> ERROR: interval field value out of range: "3000000000 days"
>
> SELECT INTERVAL '2000000000 years'
> -147483648 years
>
> I'm surprised this hasn't been reported yet. The expected behaviour is to
> raise an error.
The attached patch fixes the problem:
test=> SELECT INTERVAL '2000000000 years';
ERROR: interval out of range
LINE 1: SELECT INTERVAL '2000000000 years';
It also fixes this case
test=> SELECT INTERVAL '2000000000-3 years';
ERROR: interval field value out of range: "2000000000-3 years"
LINE 1: SELECT INTERVAL '2000000000-3 years';
--
Bruce Momjian <bruce@momjian.us> http://momjian.us
EnterpriseDB http://enterprisedb.com
+ Everyone has their own god. +
Attachment
Hello Bruce, thank you for your contribution. However, as seen from my later report, the situation is much much worse (complex, if you want) than this bug. Please see this: http://www.postgresql.org/message-id/CAMWF=HS++N9-NKsh-o5QSymvp0Np-VB0GdWwAt4uie1h8ZdNTQ@mail.gmail.com Rok Kralj, PostgreSQL enthusiast 2014-01-25 Bruce Momjian <bruce@momjian.us> > On Sun, Jun 16, 2013 at 11:03:53AM +0000, rok.kralj@gmail.com wrote: > > The following bug has been logged on the website: > > > > Bug reference: 8230 > > Logged by: Rok Kralj > > Email address: rok.kralj@gmail.com > > PostgreSQL version: 9.2.4 > > Operating system: Arch Linux > > Description: > > > > As far as I know, the INTERVAL type is represented using three int32, one > > for months, days and milliseconds. > > > > However, overflow detections seems to work correctly only on these three > > types; if they get overflown indirectly, they do not get detected: > > > > SELECT INTERVAL '3000000000 days' > > ERROR: interval field value out of range: "3000000000 days" > > > > SELECT INTERVAL '2000000000 years' > > -147483648 years > > > > I'm surprised this hasn't been reported yet. The expected behaviour is to > > raise an error. > > The attached patch fixes the problem: > > test=> SELECT INTERVAL '2000000000 years'; > ERROR: interval out of range > LINE 1: SELECT INTERVAL '2000000000 years'; > > It also fixes this case > > test=> SELECT INTERVAL '2000000000-3 years'; > ERROR: interval field value out of range: "2000000000-3 years" > LINE 1: SELECT INTERVAL '2000000000-3 years'; > > -- > Bruce Momjian <bruce@momjian.us> http://momjian.us > EnterpriseDB http://enterprisedb.com > > + Everyone has their own god. + > -- eMail: rok.kralj@gmail.com
On Sat, Jan 25, 2014 at 10:07:36PM +0100, Rok Kralj wrote: > Hello Bruce, > > thank you for your contribution. However, as seen from my later report, the > situation is much much worse (complex, if you want) than this bug. Please see > this: > > http://www.postgresql.org/message-id/CAMWF= > HS++N9-NKsh-o5QSymvp0Np-VB0GdWwAt4uie1h8ZdNTQ@mail.gmail.com Yes, I see that in my mailbox and will address those shortly, but at least this patch fixes part of the problem. -- Bruce Momjian <bruce@momjian.us> http://momjian.us EnterpriseDB http://enterprisedb.com + Everyone has their own god. +