Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[ajf]

I've seen things you people wouldn't believe. execute_strings on fire off the shoulder of Overmars. I watched object-beams glitter in the dark near the room_goto_fix. All those moments will be lost in time, like tears...in...rain. Time to die.

[MedO]

What I just found out cost me several hours and some of my sanity:

As you probably know, you can draw sprites in GM with both scaling and rotation. Generally, you would expect that if you only scale and rotate, there is one point of the sprite which never moves in screen coordinate space (the origin of the sprite). However, Game Maker decided that it would be a good idea to have different origins for rotation and scaling which are *almost* the same. Rotation is around the pixel center of the screen space draw position, while scaling is around the upper left corner the sprite's origin pixel (in sprite coordinate space). If you only scale by 1 or -1, the two points are only half a pixel apart, which makes it difficult to notice anything is wrong until some test cases are off by one pixel and you just can't figure out why...

I worked a bit on an improved system in the last days which allows you to connect sprites to defined "rigging points", so you can basically make a new hat by (mostly) defining which pixel of the hat sprite should always track the rigging point on the head of each character. Because of the "design flaw" mentioned above, doing the calculations to have the pixel positions always match up was downright agonizing. What's also painful is that it would have been really easy if GM had an actual general 2D drawing function that takes a transformation matrix instead of just rotations and scale, like any self-respecting graphics API should.

[billymaze]

I love it when Gamemaker throws syntax errors at you when you use the auto-complete drop down suggestions... retyping it out fixes the false error...

visual studio does this with c++ too

[MedO]

So since I'm already resurrecting this thread:

I want to get a sprite by name so that I can work with images that follow a consistent naming pattern (e.g. ScoutRedTauntS). However, there is no function for this, all the sprite functions work with sprite IDs. Of course, we can get around this because we basically have eval:

Code: [Select]

return execute_string("return " + spriteName + ";");
The problem here is that this causes an error if no sprite of that name exists, and of course we also don't have a way of finding that out beforehand (again, no functions for working with sprite names). However, where there is a will, there is a way:

Code: [Select]

return execute_string("variable_local_set('"+spriteName+"',-1); return " + spriteName + ";");
If the sprite exists, it will actually shadow the local variable so you get the sprite ID. Otherwise you get the contents of the local variable which is -1.

[Mr.kown]

but how wold you do it 

[Saniblues]

So since I'm already resurrecting this thread:

I want to get a sprite by name so that I can work with images that follow a consistent naming pattern (e.g. ScoutRedTauntS). However, there is no function for this, all the sprite functions work with sprite IDs. Of course, we can get around this because we basically have eval:

Code: [Select]

return execute_string("return " + spriteName + ";");
The problem here is that this causes an error if no sprite of that name exists, and of course we also don't have a way of finding that out beforehand (again, no functions for working with sprite names). However, where there is a will, there is a way:

Code: [Select]

return execute_string("variable_local_set('"+spriteName+"',-1); return " + spriteName + ";");
If the sprite exists, it will actually shadow the local variable so you get the sprite ID. Otherwise you get the contents of the local variable which is -1.

Couldn't you just write a script that checks to see if the sprite exists via sprite_exists and return a dummy sprite if it doesn't exist to the same effect? That's useful for GM8.x, but not so much for GMStudio (Not sure about GMS2 since they added the local_variable functions again, but yeah).

GM:S also has the wonderful asset_get_index(); function. Something silly like
sprite_index = asset_get_index(string(class)+string(team)+"TauntS")
is possible. We really should just drop GM8 and move on to Studio.

[MedO]

asset_get_index sounds like it would solve the problem, but sprite_exists takes an index, not a name.

As far as moving to a newer version of GM is concerned, I don't really feel like it'll help much at this point and I'd rather spend the little time I invest into GG these days on actual game or code improvements than on porting.

Besides, it's kind of fun to figure out strange ways of accomplishing things that were not intended

[Saniblues]

You got me there. Roundabout methods are the most interesting.

[ZaSpai]

As far as moving to a newer version of GM is concerned, I don't really feel like it'll help much at this point and I'd rather spend the little time I invest into GG these days on actual game or code improvements than on porting.

FWIW, even though I haven't looked into GM:S 2.0, am I still correct in assuming plugins ain't gonna work if GG2 were rebuilt in GM:S because external code cannot be executed?

[Saniblues]

As far as moving to a newer version of GM is concerned, I don't really feel like it'll help much at this point and I'd rather spend the little time I invest into GG these days on actual game or code improvements than on porting.

FWIW, even though I haven't looked into GM:S 2.0, am I still correct in assuming plugins ain't gonna work if GG2 were rebuilt in GM:S because external code cannot be executed?

This is only partially false. The reason that we were able to execute plugins the way that we do is because Game Maker pre-Studio did not compile the code. Therefore, we were able to load plain GML to the system.

http://yal.cc/on-gamemaker-studio-game-decompilation/

The method that we used for Gang Garrison 2 is no longer possible, but GMS2 adds many of the functions that would be needed to set up a modding frontend, and the functionality possibly still exists in the test versions of GMS1. YellowAfterlife, using the latter, did just this with this mod for Nuclear Throne

https://yellowafterlife.itch.io/nuclear-throne-together/

It wouldn't be a simple undertaking, and it frankly wouldn't be worth the effort in my opinion, but it's certainly not impossible.

[Saniblues]

On that note, it might also be possible with Studio as it is? I would need to ask yal if it's possible with the stable version of studio. Some of the nicer features of his modding API would be absent if it is, but hey

[Saniblues]

I assume it'll work in the next (gms) update - there's a whole bunch of memory-related bugs that I've reported while doing this and that are only fixed in closed beta as of now

[MedO]

Just found a perfect bug. It is so brilliant that it almost feels like a designed logic puzzle. Because of that, I'll first post what happened and then the solution in a spoiler.

Here is the code:

Code: [Select]

// dealDamage( sourcePlayer, damagedObject, damageDealt )
with(argument1)
{
    if(variable_local_exists("deathmatch_invulnerable"))
    {
        if(argument1.deathmatch_invulnerable != 0)
            return 0;
    }
}

argument1.hp -= argument2;

execute_string( global.dealDamageFunction, argument0, argument1, argument2 );

And here is the error, which occurs when someone is being hit directly with a rocket:

Code: [Select]

ERROR in
action number 1
of Other Event: User Defined 5
for object Rocket:

In script damageAuto:
In script damageCharacter:
In script dealDamage:
Error in code at line 6:
           if(argument1.deathmatch_invulnerable != 0)
                        ^
at position 23: Unknown variable deathmatch_invulnerable

How can this happen? We check whether the variable exists before using!
Here are some tips:
- variable_local_exists is the right function for the job (i.e. it checks if the variable exists in the "current" instance)
- There is no subtle typo involved
- argument1 has the same value inside and outside the with()-block
- This is no "weird instance state" thing where an instance is really already destroyed or something
- The with-block runs for one instance, and it's the correct one (the character which was hit by the rocket)

Maybe you'll figure it out quickly, but otherwise just check here:

(click to show/hide)

The damagedObject (argument1) that was passed in is -2, which is the same as the special value "other".

In the context this code is in (a collision event from rocket with character), that actually does resolve to the correct instance, so the with-block runs for the player who was hit. However, inside a with-block, "other" refers to the instance that was current *outside* the with-block, i.e. the rocket. So while variable_local_exists() checks for the variable in the correct instance, argument1.deathmatch_invulnerable actually tries to access the variable in the rocket instance, and fails.

[ajf]

16:13 - ΘSaniblues:

Code: [Select]

YellowAfterlife - Today at 10:12 AM
I brought it up a few times, GML uses multiple things interchangeably. For example,
it also supports pascal-style := assignment, but it's interpreted as regular assignment
token and thus you can do if (a := b) while that semantically makes no sense neither in
GML or Pascal

[Saniblues]

For the record, that's still valid as of GMS2