Use of uninitialized variables in ExecFindPartition() for parent partition without leaves (HEAD only)

From: Michael Paquier <michael(dot)paquier(at)gmail(dot)com>
To: PostgreSQL mailing lists <pgsql-hackers(at)postgresql(dot)org>
Cc: Amit Langote <amitlangote09(at)gmail(dot)com>, Robert Haas <robertmhaas(at)gmail(dot)com>
Subject: Use of uninitialized variables in ExecFindPartition() for parent partition without leaves (HEAD only)
Date: 2017-11-30 00:07:06
Message-ID: CAB7nPqQ3mwkdMoPY-ocgTpPnjd8TKOadMxdTtMLvEzF8480Zfg@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Hi all,

Since commit 4e5fe9ad (committer Robert Haas and author Amit Langote),
coverity has been complaining that the new code of ExecFindPartition()
may use a set of values and isnull values which never get initialized.
This is a state which can be easily reached with the following SQLs of
a parent partition with no children:
create table parent_tab (a int) partition by list ((a+0));
insert into parent_tab values (1);

The mistake is visibly that FormPartitionKeyDatum() should be called
even for a root partition, initializing the fields of isnull and
values on the way. That's actually what happens in execMain.c's
ExecFindPartition() if you look at REL_10_STABLE. So the answer would
be to move a bit down the quick exit code. Attached is a patch.

Amit, that's your code. What do you think?

Thanks,
--
Michael

Attachment Content-Type Size
exec-partition-quickexit.patch application/octet-stream 922 bytes

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Petr Jelinek 2017-11-30 00:30:49 Re: [HACKERS] Issues with logical replication
Previous Message Andres Freund 2017-11-29 23:47:57 Re: [HACKERS] Issues with logical replication