30 Jan, 2007, Davion wrote in the 2nd comment:
I've been looking over this for a bit trying to take it all in. I'm convinced you're getting some memory trampling somewhere. I'm basing this mostly on->
The 'modern' there is your build, and corrisponds to &w in the nanny function. I think it's reset_colors. There's a few things about this function that just scream unsafe. My main beef with this is
You have to make sure that max_colors isn't larger than the array sized for colors. Because if max_colors > sizeof ch->colors, you'll get that memory trampling and your descriptors will go bye bye.
Another reason for this is because if you're a droid reset_colors() is -never- called on you. The reason you fail with a droid is another reason. It's because of Create_object. It also doesn't crash, it's closed via exit(1); To remedy this part, make sure you look at the many calls to create_obj (It starts at line 964 in the pastebin for nanny())
I of course could be way off. Another suggestion, run under valgrind with –leak-check=full, pound at it and post output here.
# Message of the Day
# Use 'hedit motd' to write your own message of the day.
#
# Press [ENTER] modern
The 'modern' there is your build, and corrisponds to &w in the nanny function. I think it's reset_colors. There's a few things about this function that just scream unsafe. My main beef with this is
if( !str_cmp( word, "MaxColors" ) )
{
max_colors = fread_number( fp );
continue;
}
if( !str_cmp( word, "Colors" ) )
{
for( x = 0; x < max_colors; ++x )
ch->colors[x] = fread_number( fp );
continue;
}
You have to make sure that max_colors isn't larger than the array sized for colors. Because if max_colors > sizeof ch->colors, you'll get that memory trampling and your descriptors will go bye bye.
Another reason for this is because if you're a droid reset_colors() is -never- called on you. The reason you fail with a droid is another reason. It's because of Create_object. It also doesn't crash, it's closed via exit(1); To remedy this part, make sure you look at the many calls to create_obj (It starts at line 964 in the pastebin for nanny())
I of course could be way off. Another suggestion, run under valgrind with –leak-check=full, pound at it and post output here.
30 Jan, 2007, Guest wrote in the 3rd comment:
The color overflow could certainly be a problem, it was fixed here: http://www.smaugmuds.org/index.php?a=top... but I don't know if Keberus uses the exact same color code as the FOTE FUSS version. But they all derive from the color code snippet for Smaug that I've uploaded here.
Davion may well be right since the descriptor has obviously been smashed.
Quote
0 0x080fa4a9 in stop_idling (ch=0x86ad910) at comm.c:2983
2983 if( !ch || !ch->desc || ch->desc->connected != CON_PLAYING || !IS_IDLE( ch ) )
(gdb) p ch
$1 = (CHAR_DATA *) 0x86ad910
(gdb) p ch->desc
$2 = (DESCRIPTOR_DATA *) 0xa0007
(gdb) p ch->desc->connected
Cannot access memory at address 0xa0033
(gdb)
2983 if( !ch || !ch->desc || ch->desc->connected != CON_PLAYING || !IS_IDLE( ch ) )
(gdb) p ch
$1 = (CHAR_DATA *) 0x86ad910
(gdb) p ch->desc
$2 = (DESCRIPTOR_DATA *) 0xa0007
(gdb) p ch->desc->connected
Cannot access memory at address 0xa0033
(gdb)
Davion may well be right since the descriptor has obviously been smashed.
30 Jan, 2007, Kayle wrote in the 4th comment:
I'll attempt what Davion said, but as I've said on ichat many many times, I don't know how to use valgrind, but I'd really like to learn, so if someone wants to post an article about it that'd be great :tongue:
Keberus doesn't use the same color code, but I swapped it out for the same color code. which is when all this began. and now that Davion pointed out the reset_colors thing, it's all starting to make a little more sense.
Keberus doesn't use the same color code, but I swapped it out for the same color code. which is when all this began. and now that Davion pointed out the reset_colors thing, it's all starting to make a little more sense.
30 Jan, 2007, Kayle wrote in the 5th comment:
Ok, I added in the color fix, and that handled everything. Now I just have to figure out why create_object is dying.
30 Jan, 2007, Kayle wrote in the 6th comment:
Well, that was easy enough to fix, I deleted all the areas, and uploaded fresh copies and viola, create object problem vanished. Now it works perfectly. I just have to fix the horrid color output now. XD
31 Jan, 2007, Conner wrote in the 7th comment:
Gratz! :smile:
New Color.c
load_char_obj *Not sure of the relevance here, Davion wanted to see it too, so it might be related.
nanny( )
This is the nanny( ).
Now for a little additional information.
This is where the crash occurs:
This is the gdb/bt information:
Now, maybe someone here can figure this out, because I've been fighting with it all damn day trying to decipher it, or maybe Keb will even stop by and figure it out. But I've been working on this since roughly 1:30pm EST, it's now 3:40am EST and I'm handing the towel to the next person in line, while I go and try and get some sleep before Class tonight. :sad:
EDIT: Something else occured to me and I felt the need to test it. here's teh results.
Getting through as a Droid:
GDB: