Kyle Benesch

> The main dependency I used was a priority queue I believe. This is the main part. It's trivial to implement a pathfinder after you have access to a good...

Simple things like "what is the size of an index" are not so easy to resolve in C. This is one of the things which made the internal heap queue...

Thanks for sharing your own examples. A `malloc` on each queue push isn't good for performance. It harms threading and memory locality. It's tricky to figure out what I'd even...

Be sure you're using the latest alpha build. I wouldn't be surprised if some of these issues were a fixed regression. You'll need a build as close to master as...

> About the problem of "TCOD_CHAR_HLINE" and similar, aren't these commonly used for windows frames and similar? It's kind of weird that I'm the only one using these. > But...

> The example libtcod program has an "SDL callback" that in the description "gives access to the screen surface". > > The idea is that once you have access to...

The `GLSL` renderer was removed some time ago and just forwards to `OpenGL2`. So the artifacts are only an issue with the OpenGL 2 code. Different results on different machines...

If you're adding padding programmatically then you should copy the nearest pixels for the padding. With 1 pixel of padding (1 pixel each side) the shader clamping logic is no...

> The major "fix" seems to be using texture2DLod() in the fragment shader instead of texture2D(), I knew the solution could be something both simple and obscure. Thanks for figuring...

> did not make a PR because I don't feel it's ready for merge (especially with GL1 being broken), but if you are still interested in the context then I...