| Lists: | pgsql-hackers |
|---|
| From: | Kohei KaiGai <kaigai(at)kaigai(dot)gr(dot)jp> |
|---|---|
| To: | Robert Haas <robertmhaas(at)gmail(dot)com> |
| Cc: | PgHacker <pgsql-hackers(at)postgresql(dot)org> |
| Subject: | [bug] relcache leaks in get_object_address |
| Date: | 2011-10-13 15:10:34 |
| Message-ID: | CADyhKSXFm6yJpJnHNAFecNMc8eq9KpWV7q05efjRgsZ_Vubxkw@mail.gmail.com |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Lists: | pgsql-hackers |
I noticed a problem of get_object_address() with missing_ok = true.
When we try to solve the name of nonexistent rule/trigger/constraint on
a particular existing table, get_object_address_relobject() opens the
relation, but address.objectId = InvalidOid shall be set.
I think it should be closed if the queried object was missing, although
existing code does not invoke get_object_address() with missing_ok = true.
Thanks,
--
KaiGai Kohei <kaigai(at)kaigai(dot)gr(dot)jp>
| From: | Kohei KaiGai <kaigai(at)kaigai(dot)gr(dot)jp> |
|---|---|
| To: | Robert Haas <robertmhaas(at)gmail(dot)com> |
| Cc: | PgHacker <pgsql-hackers(at)postgresql(dot)org> |
| Subject: | Re: [bug] relcache leaks in get_object_address |
| Date: | 2011-10-13 16:27:12 |
| Message-ID: | CADyhKSUy4RPFumXp=-YOxm+nTejef7VPHaWtvSMsCovuh3=iiQ@mail.gmail.com |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Lists: | pgsql-hackers |
The attached patch fixes this problem.
Unfortunately, we have no code that invokes get_object_address()
with missing_ok = true now, so please apply a couple of patches
to rework DROP statement of mine.
DROP TRIGGER no_such_trigger ON existing_table;
shall cause a relcache reference leaks, without this patch.
Thanks,
2011/10/13 Kohei KaiGai <kaigai(at)kaigai(dot)gr(dot)jp>:
> I noticed a problem of get_object_address() with missing_ok = true.
>
> When we try to solve the name of nonexistent rule/trigger/constraint on
> a particular existing table, get_object_address_relobject() opens the
> relation, but address.objectId = InvalidOid shall be set.
>
> I think it should be closed if the queried object was missing, although
> existing code does not invoke get_object_address() with missing_ok = true.
>
> Thanks,
> --
> KaiGai Kohei <kaigai(at)kaigai(dot)gr(dot)jp>
>
--
KaiGai Kohei <kaigai(at)kaigai(dot)gr(dot)jp>
| Attachment | Content-Type | Size |
|---|---|---|
| pgsql-v9.2-fix-relcache-leaks-in-get_object_address_relobject.patch | application/octet-stream | 953 bytes |
| From: | Robert Haas <robertmhaas(at)gmail(dot)com> |
|---|---|
| To: | Kohei KaiGai <kaigai(at)kaigai(dot)gr(dot)jp> |
| Cc: | PgHacker <pgsql-hackers(at)postgresql(dot)org> |
| Subject: | Re: [bug] relcache leaks in get_object_address |
| Date: | 2011-10-14 15:41:25 |
| Message-ID: | CA+TgmoZ30Cr-s02h97vz=8v+vj-FdTtWu6hOAyzBe3w-YopwvA@mail.gmail.com |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Lists: | pgsql-hackers |
On Thu, Oct 13, 2011 at 12:27 PM, Kohei KaiGai <kaigai(at)kaigai(dot)gr(dot)jp> wrote:
> The attached patch fixes this problem.
> Unfortunately, we have no code that invokes get_object_address()
> with missing_ok = true now, so please apply a couple of patches
> to rework DROP statement of mine.
>
> DROP TRIGGER no_such_trigger ON existing_table;
>
> shall cause a relcache reference leaks, without this patch.
We could argue that that's a bug in your drop reworks rather than the
existing code, but I'm willing to go along with this approach, so,
committed.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company