OK, this is a weird one... we wanted to be able to reload the fight messages,
so they could be changed and tested immediately. I didn't want a big memory
leak, so, I changed load_messages() as follows and added a call to it in
do_reboot for "reload messages":
void load_messages(void)
{
FILE *fl;
int i, type;
struct message_type *messages, *next;
char chk[128];
if (!(fl = fopen(MESS_FILE, "r"))) {
sprintf(buf2, "Error reading combat message file %s", MESS_FILE);
perror(buf2);
exit(1);
}
for (i = 0; i < MAX_MESSAGES; i++) {
fight_messages[i].a_type = 0;
fight_messages[i].number_of_attacks = 0;
messages = fight_messages[i].msg;
while (messages) {
if (messages->die_msg.attacker_msg)
free(messages->die_msg.attacker_msg);
if (messages->die_msg.victim_msg)
free(messages->die_msg.victim_msg);
if (messages->die_msg.room_msg)
free(messages->die_msg.room_msg);
if (messages->miss_msg.attacker_msg)
free(messages->miss_msg.attacker_msg);
if (messages->miss_msg.victim_msg)
free(messages->miss_msg.victim_msg);
if (messages->miss_msg.room_msg)
free(messages->miss_msg.room_msg);
if (messages->hit_msg.attacker_msg)
free(messages->hit_msg.attacker_msg);
if (messages->hit_msg.victim_msg)
free(messages->hit_msg.victim_msg);
if (messages->hit_msg.room_msg)
free(messages->hit_msg.room_msg);
if (messages->god_msg.attacker_msg)
free(messages->god_msg.attacker_msg);
if (messages->god_msg.victim_msg)
free(messages->god_msg.victim_msg); /* 2 */
if (messages->god_msg.room_msg)
free(messages->god_msg.room_msg); /* 1 */
next = messages->next;
free(messages);
messages = next;
}
fight_messages[i].msg = 0;
}
... stock code to load the messages ...
It crashed with a segfault in libc_free on the line marked "1", when attempting
to free the last message loaded. I determined this by stepping through with
gdb. The message was indeed allocated and all seemed fine, no reason for it to
crash. So, I went into the messages file and made the last message a "#" (no
message). Now it crashes on the line marked "2". I'm completely befuddled. I
added a check to both of the marked lines by checking for the the last message
and not freeing it (&& fight_messages[i+1].a_type !=0), and sure enough, the
crash went away, but we now have a memory leak. There's nothing unusual about
the way the last message is read, since there is an extra line on the messages
file ("$"). It doesn't know it's gonna be the last line when it reads it, so
nothing is done differently.
It's a Pentium system running Linux, not sure what library levels and such, but
they aren't very old.
Any ideas?
Ron
+------------------------------------------------------------+
| Ensure that you have read the CircleMUD Mailing List FAQ: |
| http://democracy.queensu.ca/~fletcher/Circle/list-faq.html |
+------------------------------------------------------------+
This archive was generated by hypermail 2b30 : 12/08/00 PST