search

zturtleman / spearmint

Spearmint — an updated id Tech 3 engine for continuing the classics and creating new games.
https://clover.moe/spearmint
GNU General Public License v3.0
237 stars 27 forks source link

Another bot issue in 17t6remix/17t6ctf #223

Closed rawr51919 closed 8 years ago

rawr51919 commented 8 years ago

There happens to be an issue with botplay in these 2 maps in which they get stuck in one portion of the floor (it's more evident in 17t6ctf). Happens even with the latest spearmint. ioquake3 does not seem to have this error. 17t6remix link: http://lvlworld.com/review/id:1847 17t6ctf link: https://ws.q3df.org/map/17t6ctf/ Botplay patch link: https://mega.nz/#!F4tAmChI!oMWbHT6ZJBt2NbFJ70uhAjfMXZUbrbIpyAgUsu-mRWc The reason for the patch is they wouldn't do anything without you shooting at them in both maps. With the patch, they'll go almost everywhere, but they're not able to grab the MH or LG for some reason.

ghost commented 8 years ago

I did a quick check about this issue: AFAIK, this is not directly related to AI game/engine code. The issue you encountered depends on the AAS file for those maps. The AAS file is the file bots use to navigate through the bsp world. In some cases the map is built badly, this causes the bots to get stuck in 'solid areas'. You can test this by starting a map in developer mode (/devmap mapname), and use the 'bot_testsolid' command, entered into your console after the map is loaded. If the map has some 'solid areas' the console will print 'SOLID area' on the top left corner. See the screenshot below (taken from 17t6remix). 17remix

Your patch from above simply holds different AAS files (eventually compiled with different AAS switches). NOTE: Care must be taken when creating AAS files for maps that were compiled with the -meta switch during the BSP Phase, in this case you MUST compile AAS files using the 'forcesidesvisible' switch, otherwisse the AAS file will always be 'solid' (or not created at all). Some more basic instructions about 'Testing AAS files': https://icculus.org/gtkradiant/documentation/Compile_Manual/bspc.txt

rawr51919 commented 8 years ago

Never thought about that... I'll try & see if I can fix that with another aas file (forcesidesvisible was already used).

On 5/5/16, ToKu notifications@github.com wrote:

I did a quick check about this issue: AFAIK, this is not directly related to AI game/engine code. The issue you encountered depends on the AAS file for those maps. The AAS file is the file bots use to navigate through the bsp world. In some cases the map is built badly, this causes the bots to get stuck in 'solid areas'. You can test this by starting a map in developer mode (/devmap mapname), and use the 'bot_testsolid' command, entered into your console after the map is loaded. If the map has spme 'solid areas' the console will print 'SOLID AREA' on your top left Corner. See the screenshot below (taken from 17t6remix). 17remix

Your patch from above simply holds different AAS files. NOTE: Care must be taken when creating AAS files for maps that were compiled with the -meta switch during the BSP Phase, in this case you MUST compile AAS files using the 'forcesidesvisible' Switch, otherwisse the AAS file will always be 'solid' (or not created at all). Some more Basic instructions about 'Testing AAS files': https://icculus.org/gtkradiant/documentation/Compile_Manual/bspc.txt

You are receiving this because you authored the thread. Reply to this email directly or view it on GitHub: https://github.com/zturtleman/spearmint/issues/223#issuecomment-217114582

rawr51919 commented 8 years ago

It doesn't seem to fix this with BSPC 2.1i. 2.1j might work. I'll also try 2.1h & 1.7 to see if they work as well. EDIT: 1.7 doesn't even support forcesidesvisible. I'll try 2.0 as well.

zturtleman commented 8 years ago

You could try BSPC in the Spearmint repo as well. make -C code/bspc. Note that force sides visible and grapple reach are enabled by default.

rawr51919 commented 8 years ago

Maybe you might want to put the make command for bspc in the readme.md on github zturtleman.

Sent from my iPad

On May 5, 2016, at 12:10 PM, Zack Middleton notifications@github.com wrote:

You could try BSPC in the Spearmint repo as well. make -C code/bspc. Note that force sides visible and grapple reach are enabled by default.

— You are receiving this because you authored the thread. Reply to this email directly or view it on GitHub

rawr51919 commented 8 years ago

Spearmint BSPC does not fix this problem...

ghost commented 8 years ago

I'm in doubt you'll fix this issue by using different bspc versions. Remember, the problem is the map file, from which the bsp file was created from, NOT bspc or even game code. Note: the review to this map already mentions that bots won't work properly: http://lvlworld.com/review/id:1847 So, you need the source file of the map to fix this issue..., the map file is owned by the author of the map.

rawr51919 commented 8 years ago

Either that, or the map can be decompiled and fixed just as easily using the q3map2 from gtkradiant 1.6.5. 1.6.5's gtkradiant has texture alignment support (to a degree).

On 5/5/16, ToKu notifications@github.com wrote:

I'm in doubt you'll fix this issue by using different bspc versions. Remember, the problem is the map file, from which the bsp file was created from, NOT bspc or even game code. Note: the review to this map already mentions that bots won't work properly: http://lvlworld.com/review/id:1847 So, you need the source file of the map to fix this issue..., the map file is owned by the author of the map.

You are receiving this because you authored the thread. Reply to this email directly or view it on GitHub: https://github.com/zturtleman/spearmint/issues/223#issuecomment-217292907

ensiform commented 8 years ago

Decompiling the map trashes lighting, texture alignment and sometimes incorrect brushwork. Its more work than you think. And besides its not like its necessarily "legal". The map's design in general is quite poor I think is the issue, no?

rawr51919 commented 8 years ago

Um, no. The latest q3map2 keeps lighting, brushwork, and texture alignment (to a degree), as I specified earlier. I tried this myself, and I was able to compile a decompiled map with lighting, brushwork, and texture alignment all virtually the same as the original. Try it for yourself and PROVE IT that I'm wrong.

On Thursday, 5 May 2016, Ensiform notifications@github.com wrote:

Decompiling the map trashes lighting, texture alignment and sometimes incorrect brushwork. Its more work than you think. And besides its not like its necessarily "legal". The map's design in general is quite poor I think is the issue, no?

— You are receiving this because you authored the thread. Reply to this email directly or view it on GitHub https://github.com/zturtleman/spearmint/issues/223#issuecomment-217319017

ensiform commented 8 years ago

The 1.6.5 version isn't listed anywhere that I saw. And point lights (classname "light" entity) cannot possibly be saved because they are removed from the bsp during compile unless flagged to not which most people don't do. Shader lights of course will be there because the shaders are there when you recompile.

rawr51919 commented 8 years ago

It's on Quake3World. Just Google search "gtkradiant 1.6.5" & you'll find it.

On Thursday, 5 May 2016, Ensiform notifications@github.com wrote:

The 1.6.5 version isn't listed anywhere that I saw. And point lights (classname "light" entity) cannot possibly be saved because they are removed from the bsp during compile unless flagged to not which most people don't do. Shader lights of course will be there because the shaders are there when you recompile.

— You are receiving this because you authored the thread. Reply to this email directly or view it on GitHub https://github.com/zturtleman/spearmint/issues/223#issuecomment-217322482

aisouard commented 8 years ago

I tried this myself, and I was able to compile a decompiled map with lighting, brushwork, and texture alignment all virtually the same as the original. Try it for yourself and PROVE IT that I'm wrong.

Please provide us the map you've successfully decompiled so we can take a look on it, I agree for the texture alignment since I remember using a q3map2 version coming from NetRadiant.

Regarding the lights, it's as @ensiform said, q3map2 removes the lights entities unless the author sets the parameter "_keeplights" to 1 inside the worldspawn entity.

Mateos81 commented 8 years ago

Latest GtkRadiant: https://s3.amazonaws.com/gtkradiant/GtkRadiant-1.6.5-20160424.zip

mecwerks commented 8 years ago

I think this issue should be closed due to it being a map specific bug. Issues should pertain to the engine itself and not a faulty map.