mysql_fetch_lengths and rows

[Edward Diener]

Does the 'unsigned long *' returned from a call to mysql_fetch_lengths(MYSQL_RES *result) change for each row read using mysql_fetch_row(MYSQL_RES *result) ? I would have though that it must, since the actual column lengths of data may change for each row read.

Yet when I compiled a program using gcc's -O3 optimization and debugged the program with gdb because of a program hang, I noticed that gdb told me that the unsigned long * variable I was using to hold the result from mysql_fetch_lengths had been optimized away. In other words, in a loop such as:

#include <mysql/mysql.h>
MYSQL_RES* res;
// assume res has been correctly set in the code here
MYSQL_ROW row;
unsigned long * col_lengths;
while ((row = mysql_fetch_row(res)) != 0)
{
col_lengths = mysql_fetch_lengths(res);
... more code
}

the gdb debugger tells me, when I try to 'print col_lengths' at the beginner of debugging the ... more code section, that col_lengths has been optimized away. This implies that gcc with the -O3 optimization options has decided that col_lengths is a hardcoded value that can not change in the loop.

Does anybody know that is going on here ? It sounds to me that I can not use gcc's optimization levels when calling mysql client code. Is this somehwere in the MySql docs, or have I missed something here ?

Compiling with -O0 works fine, but -O3 does not as evidenced by this problem. This is a Linux apllication using gcc-7.2 on a Debian ARM system. The program is duly linking with the mysqlclient shared library.