Please enable fullmap
"> gimme fullmap"
and try it again, looking for where they are going.
They are not dying, they are moving to someplace else.
Please see if they are moving through the back wall into the adjacent void, or just teleporting randomly out into the map.
On another map, I have seen pinkies show up outside the walls, and I have not figured out yet how they are getting loose. It seems to get pinkies more, probably because they pretty much run around bumping into walls alot.
They do not disappear on the map when I run 1.48.6 on Linux.
Let me know if it is not repeatable.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I have looked at the video several times, and it looks like they are walking through the walls in the center.
If you look at the map with "fullmap" are they piling up on the other side of that wall ?
I still cannot reproduce this on my DoomLegacy copies.
I am going to have to review all the patches over the last year, to see if any of them touched the code for wall collision detection. Years ago, every once in a while I would find a monster outside the walls in the void space.
The question is why does your machine have it so visible on map08, and mine does not.
Is your machine unusual in some way ?
Are you running with any unusual settings, command line switches ?
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
A blockmap problem would make it possible for monsters to go through some walls.
Do you run that copy of doom2 with any other Doom ports, ones that make their own blockmap ?
Can you verify that the blockmap on that copy of Doom2 has not been damaged ?
Can you do a compare against a known good copy (using diff or cmp) ?
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Monsters do not go through the wall, they teleport outside the map. Later they appear from the left side at time interval which is not the same as they disappear. (new video with gimme fullmap attached) It seems they teleport to different places at different distance so it takes long time for them to return. I watched this bug on other maps as well. As i remember, for map07 spiders appear far away outside the walls of the level, i cant kill them all and pass to next level. My machine is absolutely common, fresh native windows installation (real, not virtualized), a download of doomlegacy and its commons with all default settings except video. I tried doom2.wad (14 604 584 bytes) from different sources, the bug is still there. The error is consistently reproduced, not occasionally. I'm surprised you do not have it.
I have had a similar problem on TNT map 22, where one or two of the pinkies from the final area to the north will start showing up from the void to the south, trying to get to me through the wall.
But today it will not do it. That makes it hard to debug this on my machine.
It probably has something to do with the layout of allocations in Z_Malloc, or some other user or machine dependent alignments.
Your pinkies are coming from the left, but mine come from the south.
You seem to have the most reproducible occurance.
I will likely have to write a special debug version of the code that watches all the monsters and tries to log what code was messing with them at the time this bug hits. This will take a while.
Do you compile code, or just run pre-compiled binaries ?
It is likely that every binary changes where this bug occurs, so they cannot be directly compared. But if you have had any older binaries, like 1.47.2, it would be useful to know the first binary that shows this bug.
I have copies of the binary (Linux) for every patch, but cannot get the bug to show itself reliably.
This will take a while so do not expect to hear anything soon, and I have another sky bug to fry right now. If you find out anything new, please let me know. Thanks.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
This was caused by a broken calculation to determine the opposite direction.
This bug was introduced in the moonwalk patch (svn 1540) when I copied
the opposite calculation code from PrBoom, in an effort to reduce
differences that might lead to bugs, replacing a table lookup
implementation. However, the PrBoom calculation must be guarded against
DI_NODIR, which the table implementation could handle inherently.
Without that protection, an olddir of DI_NODIR introduced a direction
of 12 into the logic, which can only handle directions of 0..7, with
NODIR=8. Several times a game that 12 value would survive long enough
to get used, which would be expressed as a wild walk movement.
Both options were fixed, with the fixed calculation barely winning a
speed comparison over the table lookup, and having a slightly smaller
cache impact.
This affects all game play enough that a new release will be scheduled soon.
For the time being, on some levels, monsters will escape and the player will not be able to get 100% kills.
Because I work on Linux, I cannot get a Windows test copy of the new code to you for testing.
The new code is in the SVN system, for anyone who can compile their own.
Thank you for the help. I had noticed the loose monsters on TNT map22, but this bug had mostly escaped detection during testing.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Please enable fullmap
"> gimme fullmap"
and try it again, looking for where they are going.
They are not dying, they are moving to someplace else.
Please see if they are moving through the back wall into the adjacent void, or just teleporting randomly out into the map.
On another map, I have seen pinkies show up outside the walls, and I have not figured out yet how they are getting loose. It seems to get pinkies more, probably because they pretty much run around bumping into walls alot.
They do not disappear on the map when I run 1.48.6 on Linux.
Let me know if it is not repeatable.
I have looked at the video several times, and it looks like they are walking through the walls in the center.
If you look at the map with "fullmap" are they piling up on the other side of that wall ?
I still cannot reproduce this on my DoomLegacy copies.
I am going to have to review all the patches over the last year, to see if any of them touched the code for wall collision detection. Years ago, every once in a while I would find a monster outside the walls in the void space.
The question is why does your machine have it so visible on map08, and mine does not.
Is your machine unusual in some way ?
Are you running with any unusual settings, command line switches ?
A blockmap problem would make it possible for monsters to go through some walls.
Do you run that copy of doom2 with any other Doom ports, ones that make their own blockmap ?
Can you verify that the blockmap on that copy of Doom2 has not been damaged ?
Can you do a compare against a known good copy (using diff or cmp) ?
Monsters do not go through the wall, they teleport outside the map. Later they appear from the left side at time interval which is not the same as they disappear. (new video with gimme fullmap attached) It seems they teleport to different places at different distance so it takes long time for them to return. I watched this bug on other maps as well. As i remember, for map07 spiders appear far away outside the walls of the level, i cant kill them all and pass to next level. My machine is absolutely common, fresh native windows installation (real, not virtualized), a download of doomlegacy and its commons with all default settings except video. I tried doom2.wad (14 604 584 bytes) from different sources, the bug is still there. The error is consistently reproduced, not occasionally. I'm surprised you do not have it.
I have had a similar problem on TNT map 22, where one or two of the pinkies from the final area to the north will start showing up from the void to the south, trying to get to me through the wall.
But today it will not do it. That makes it hard to debug this on my machine.
It probably has something to do with the layout of allocations in Z_Malloc, or some other user or machine dependent alignments.
Your pinkies are coming from the left, but mine come from the south.
You seem to have the most reproducible occurance.
I will likely have to write a special debug version of the code that watches all the monsters and tries to log what code was messing with them at the time this bug hits. This will take a while.
Do you compile code, or just run pre-compiled binaries ?
It is likely that every binary changes where this bug occurs, so they cannot be directly compared. But if you have had any older binaries, like 1.47.2, it would be useful to know the first binary that shows this bug.
I have copies of the binary (Linux) for every patch, but cannot get the bug to show itself reliably.
This will take a while so do not expect to hear anything soon, and I have another sky bug to fry right now. If you find out anything new, please let me know. Thanks.
I tried 1.48.4 rev1531 for win as well. Everything works fine, no bug there. I do not compile code, just run binaries from https://sourceforge.net/projects/doomlegacy/files/
Fixed SVN 1563.
This was caused by a broken calculation to determine the opposite direction.
This bug was introduced in the moonwalk patch (svn 1540) when I copied
the opposite calculation code from PrBoom, in an effort to reduce
differences that might lead to bugs, replacing a table lookup
implementation. However, the PrBoom calculation must be guarded against
DI_NODIR, which the table implementation could handle inherently.
Without that protection, an olddir of DI_NODIR introduced a direction
of 12 into the logic, which can only handle directions of 0..7, with
NODIR=8. Several times a game that 12 value would survive long enough
to get used, which would be expressed as a wild walk movement.
Both options were fixed, with the fixed calculation barely winning a
speed comparison over the table lookup, and having a slightly smaller
cache impact.
This affects all game play enough that a new release will be scheduled soon.
For the time being, on some levels, monsters will escape and the player will not be able to get 100% kills.
Because I work on Linux, I cannot get a Windows test copy of the new code to you for testing.
The new code is in the SVN system, for anyone who can compile their own.
Thank you for the help. I had noticed the loose monsters on TNT map22, but this bug had mostly escaped detection during testing.