Message-ID: <f3029f25-acc9-4cb9-a74f-fe93bcfb3a27@gmail.com>
Date: Sat, 7 Mar 2026 03:28:14 +0300
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Content-Language: en-US, ru
To: pgsql-hackers@postgresql.org
From: George Tarasov <george.v.tarasov@gmail.com>
Subject: [BUG?] missing array index may result in a wrong constraint name
 (pg_dump, bin-upgrade, >=18)
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
Archived-At: 
 <https://www.postgresql.org/message-id/f3029f25-acc9-4cb9-a74f-fe93bcfb3a27%40gmail.com>
Precedence: bulk

     Greetings!

I performed static code analysis and got an error: ALWAYS FALSE SUBEXPR!
Let's look at a little snippet (pg_dump.c:10203).

     /*
      * In binary upgrade of inheritance child tables, must have a
      * constraint name that we can UPDATE later; same if there's a
      * comment on the constraint.
      */
     if ((dopt->binary_upgrade &&
          !tbinfo->ispartition &&
          !tbinfo->notnull_islocal) ||
          ^^^^^^^^^^^^^^^^^^^^^^^^            ALWAYS FALSE SUBEXPR!
         !PQgetisnull(res, r, i_notnull_comment))
     {

It seems like index "[j]" is missing for this subexpression to make any 
sense.

"notnull_islocal" is an array of bools.  It is _always_ created before 
this snippet
in function "getTableAttrs()".  So, the value of tbinfo->notnull_islocal 
is _always_ TRUE
(it is a valid memory pointer), and therefore the entire subexpression 
is _always_ FALSE.
And so, in binary upgrade mode this condition may not be evaluated as 
expected
(when there's NO comment on the constraint!).

To be honest, I don't fully understand the overall logic behind these 
preparations
for further processing (in pg_dump), but it seems like "[j]" index is 
definitely missing.

That is how it should be.

      /*
      * In binary upgrade of inheritance child tables, must have a
      * constraint name that we can UPDATE later; same if there's a
      * comment on the constraint.
      */
     if ((dopt->binary_upgrade &&
          !tbinfo->ispartition &&
          !tbinfo->notnull_islocal[j]) ||
                                  ^^^                  MISSING ARRAY INDEX!
         !PQgetisnull(res, r, i_notnull_comment))
     {

Dear, Hackers! Please review.
Is there a really mistake here or am I wrong to raise this issue?

Thanks!
George Tarasov