BUG #16457: Implicit signed integer truncation or sign change at sortsupport.c:60

Started by PG Bug reporting formalmost 6 years ago1 messagesbugs
Jump to latest
#1PG Bug reporting form
noreply@postgresql.org

The following bug has been logged on the website:

Bug reference: 16457
Logged by: ChiJin Zhou
Email address: tlock.chijin@gmail.com
PostgreSQL version: 12.3
Operating system: ubuntu 18.04
Description:

REPRODUCTION && ERROR MESSAGE:

compile the postgres project with UndefinedBehaviorSanitizer and start a
start a server with "postgress" (command: "./postgres -Ddata -p23333"), and
Ubsan reported an error message:
sortsupport.c:60:9: runtime error: implicit conversion from type 'Datum'
(aka 'unsigned long') of value 18446744073709551615 (64-bit, unsigned) to
type 'int' changed the value to -1 (32-bit, signed)
#0 0xf3828d in comparison_shim
/root/chijin_workshop/postgres-new/postgres_source_enhanced/src/backend/utils/sort/sortsupport.c:60:9

#1 0x20a2793 in ApplySortComparator.5353
/root/chijin_workshop/postgres-new/postgres_source_enhanced/src/backend/commands/../../../src/include/utils/sortsupport.h:224:13

#2 0x20a1c49 in compare_scalars
/root/chijin_workshop/postgres-new/postgres_source_enhanced/src/backend/commands/analyze.c:2711:12

#3 0xa9a6bb in med3.26971
/root/chijin_workshop/postgres-new/postgres_source_enhanced/src/port/qsort_arg.c:108:4

#4 0xa98d65 in qsort_arg
/root/chijin_workshop/postgres-new/postgres_source_enhanced/src/port/qsort_arg.c:158:9

#5 0x2094869 in compute_scalar_stats
/root/chijin_workshop/postgres-new/postgres_source_enhanced/src/backend/commands/analyze.c:2284:3

#6 0x2062fb6 in do_analyze_rel
/root/chijin_workshop/postgres-new/postgres_source_enhanced/src/backend/commands/analyze.c:541:4

#7 0x2053891 in analyze_rel
/root/chijin_workshop/postgres-new/postgres_source_enhanced/src/backend/commands/analyze.c:260:3

#8 0x204be25 in vacuum
/root/chijin_workshop/postgres-new/postgres_source_enhanced/src/backend/commands/vacuum.c:413:5

SUMMARY: UndefinedBehaviorSanitizer:
implicit-signed-integer-truncation-or-sign-change sortsupport.c:60:9

ROOT CAUSE CODE

The root cause code and explanations are shown bellow (sortsupport.c:60).

static int
comparison_shim(Datum x, Datum y, SortSupport ssup)
{
SortShimExtra *extra = (SortShimExtra *) ssup->ssup_extra;
Datum result;

extra->fcinfo.args[0].value = x;
extra->fcinfo.args[1].value = y;

/* just for paranoia's sake, we reset isnull each time */
extra->fcinfo.isnull = false;

result = FunctionCallInvoke(&extra->fcinfo);
<---------- the result value is usigned 64 bit

/* Check for null result, since caller is clearly not expecting one
*/
if (extra->fcinfo.isnull)
elog(ERROR, "function %u returned NULL",
extra->flinfo.fn_oid);

return result; <----------- the return value is
signed 32 bit
}