Re: BUG #17909: CREATE SCHEMA AUTHORIZATION sch CREATE TABLE foo ( id INT ) will coredump

Richard Guo <guofenglinux@gmail.com>

From: Richard Guo <guofenglinux@gmail.com>
To: Michael Paquier <michael@paquier.xyz>
Cc: hysong0101@163.com, pgsql-bugs@lists.postgresql.org
Date: 2023-04-27T08:59:13Z
Lists: pgsql-bugs
On Thu, Apr 27, 2023 at 3:34 PM Michael Paquier <michael@paquier.xyz> wrote:

> It took me a couple of minutes to get what you meant here.  The point
> is that schema-qualifying any of the object specified after the CREATE
> SCHEMA with a schema name different than the rolname would cause a
> crash, when no schema is directly given.  We should fail with the same
> error than when a schema is specified, as of, except that the rolename
> needs to be specified:
> =# create schema popo authorization postgres create table lala.aa (a int);
> ERROR:  42P15: CREATE specifies a schema (lala) different from the one
> being created (popo)


Aha, now I get the scenario that would crash.

# create schema authorization postgres create table lala.aa (a int);
server closed the connection unexpectedly

In this case the CreateSchemaStmtContext.schemaname is NULL since it is
not explicitly specified, while the schemaname in the schema element is
not NULL as it is specified, and setSchemaName cannot copy with such
situation.  Maybe we should check against RoleSpec.rolename in this case
since that is also the schema's name?


> That seems quite old, at quick glance (v11 fails), so this needs to be
> fixed all the way down.


Yes.  I can see this crash from master all back to v9.5.

Thanks
Richard

Commits

  1. Fix crashes with CREATE SCHEMA AUTHORIZATION and schema elements