shithead king poster
Working on YoGUI. I added FilterTextFields--that is, you supply a regex, and it changes its border color if it's a match or not. It also provides a boolean method that lets the programmer see if the text is validated. You can customize the success and fail border color, so if you don't want it to change colors, set them both to the same thing.
In the following pictures, the regex I'm using is:
And here it is put into Minor Destruction:
my app used to load images and then convert them to the proper format and size (i.e. pow2) at run-time
then i thought: why the fuck is it doing that as i'm the only one that tells it what images to load
so i made a little program that uses freeimage to read my artist's PNG/TGA files and outputs custom images
these custom images are quite literally
4 bytes: original width
4 bytes: original height
4 bytes: pow2 width
4 bytes: pow2 height
1 byte: RGB or RGBA pixels
N bytes: pixel data ready to be fed to opengl es
loading times? instantaneous. memory problems (running out of memory while converting, etc.)? gone.
i'm happy :)
Content wise, finished up the Reverse Key Finder in its first state! Now to add a lot more chord types and then get it working on the Android... not looking forward to that.
So this is what I'm up to.
Fixing saving in Half-Life 2: Sandbox, which is rooted in a single segfault. Not bad if you ask me; it's better than having to rewrite a bunch of shit to work with multiplayer. I'm sorta getting sick of that. Not to mention some things are plainly broken upon the logical transition from a single player interacting with various parts of a game designed for a single player to multiple players doing the same thing.
My issue is that the saverestore system in Source traverses entity hierarchies and the CHL2MP_Player class wasn't explicitly designed to be saved, so it has a "null" data description for what it needs to pass on to this system.
That's sort of where I'm stuck. Having an empty data description array isn't bad, nor is that really the issue - it seems here that Source attempts to save pointers to various think context functions if available shortly after grabbing this data description and all of its base classes', but inevitably it attempts to handle a bad function pointer to one of them, and crashes later down the execution chain.
So I've been fiddling with the debugger and animating my way down the call stack, but I haven't really hit on anything conclusive yet.
It's not content as much as it is technical insight on what I'm doing, but I am actively working on things like this every single day as well as implementing new features into the mod.
reading unaligned memory, masking off bits and then shifting 1 by that is a lot more effort for no gain rather than simply reading a uint32_t
Got some A* up in this bitch:
I'm obsessed with the idea of using volumentric analysis in 3d to plot out nodes using octrees to check if points in the cube is within the world or not(and from that decide the next action, plot node, subdivide more, etc), and WHILE doing that, not after, pathfind to your target.
I have a semi working algorithm in my head. But holy fuck I have to do this school project and I don't have the time for this shit.
Has anyone done something similar?
If not I'll gladly change it to "Graph path finding" instead. It does what I want it to do anyway
Figured out what was wrong with my GC that was causing things to be collected when they shouldn't be.
Here's the definition of the 'setindex' instruction:
Looks alright, yeah?
Nope! Basically I'm taking a pointer to the middle of a struct and saving that pointer in a hash table. When it comes time for the GC to scan for unreferenced objects, it goes and collects the struct containing the thing I'm pointing to, because there aren't any pointers pointing to that struct!
Here's what I should have been doing:
This sort of gives me the shits because I had considered this very issue a while ago and I wrote the js_to_js_string_t() function to deal specifically with this sort of thing... and then I forgot about it and did it the shit way anyway.
Here's what js_to_js_string_t() does:
Did some more work on my Lorenz Attractor visualizer. Increased the time it draws for from 1024 to 20,000, added colour changing in, and made it use a line renderer instead of GL calls. It looks much better now.
Watch in HD if you want to see it best.
Oh wow, it actually works at full speed on my phone too.
i love the glow effect and how things get brighter towards the center, looks quite cool
Gesture menu actually does something now - also pauses the game!
Can someone explain why reading unaligned memory is slow, I've seen it talked about loads but never really explained.
I've even seen some C++ structs which had 3 floats, then an extra float just for padding, supposedly because of this mysterious memory alignment.
I can see that getting extremely annoying when your moving your mouse to the right since it would get under it.