Thread: BUG #18848: DEREF_AFTER_NULL.EX.COND After having been compared to a NULL

The following bug has been logged on the website:

Bug reference:      18848
Logged by:          Nikita
Email address:      pm91.arapov@gmail.com
PostgreSQL version: 16.6
Operating system:   ubuntu 20.04
Description:

Looks like there is inconsistency with 'ind_name' and 'int_type' checks. In
one place both are checked. In other place only the second one is checked.
It's possible that when the second is not NULL the first is also not NULL.
If this is so then the problem is not real. But it's hard to prove this. In
any case the code needs to be made more clear.

diff --git a/src/interfaces/ecpg/preproc/type.c
b/src/interfaces/ecpg/preproc/type.c
--- a/src/interfaces/ecpg/preproc/type.c    (revision
a49ac80219c6f28c3cf3973f797de637329952da)
+++ b/src/interfaces/ecpg/preproc/type.c    (date 1740396929346)
@@ -395,7 +395,10 @@
 
                 ECPGdump_a_simple(o, name, type->type, type->size, (arr_str_size &&
strcmp(arr_str_size, "0") != 0) ? arr_str_size : str_neg_one, struct_sizeof,
prefix, type->counter);
                 if (ind_type != NULL)
-                    ECPGdump_a_simple(o, ind_name, ind_type->type, ind_type->size,
(arr_str_size && strcmp(arr_str_size, "0") != 0) ? arr_str_size :
ind_type_neg_one, ind_struct_sizeof, ind_prefix, 0);
+                {
+                    Assert(ind_name != NULL);
+                    ECPGdump_a_simple(o, ind_name, ind_type->type,
ind_type->size, (arr_str_size && strcmp(arr_str_size, "0") != 0) ?
arr_str_size : ind_type_neg_one, ind_struct_sizeof, ind_prefix, 0);
+                }
 
                 free(str_neg_one);
                 free(ind_type_neg_one);

PG Bug reporting form <noreply@postgresql.org> writes:
> Looks like there is inconsistency with 'ind_name' and 'int_type' checks. In
> one place both are checked. In other place only the second one is checked.
> It's possible that when the second is not NULL the first is also not NULL.
> If this is so then the problem is not real. But it's hard to prove this. In
> any case the code needs to be made more clear.

Perhaps it needs to be made more clear, but this patch seems like no
improvement.  Adding that Assert accomplishes nothing --- the code
will dump core if ind_name is null, with or without an Assert.

Given the lack of field complaints, I'm inclined to believe there's
not a bug here.

            regards, tom lane