When `i` starts with `0`, `t` is also `0`, which results in `previous == startPos == current`, this segment is not only redundant, but it also causes division-by-zero since `sqrtf(dx*dx + dy*dy)` is zero.
* Fix GetCodepointNext to return default value with size=0 on invalid input. Modify LoadCodepoints to work when GetCodepointNext returns a size of 0. All internal use of GetCodepointNext and GetCodepointPrev checked. This fix may break external code dealing with invalid input as the old code erroneously never returned a size of 0, external code that doesn't properly check for size=0 may endlessly loop or overflow a buffer on invalid input.
* Change default behaviour of GetCodepointNext to return a size of 1 instead of 0. This matches existing prod behaviour and guarantees size 1..4 is returned. Simplify internal code that uses GetCodepointNext that previously had to account for size=0.
* Simplified progressing through a UTF-8 string in ImageTextEx and MeasureTextEx. This change matches existing precedent in DrawTextEx
* GetCodepointNext: Add 10xxxxxx checks to multibyte encodings.
---------
Co-authored-by: anon <anon>
Basic packing algorithm currently follows this order: Copy pixel data -> Move offsetX for current glyph -> Check remaining space for current glyph...
Since X offset already moved according current glyph, remaining space should be checked for next glyph. Because of this, occasionally, current logic causes glyphs wrapping around texture.
Proposed fix accomplishes that by moving offsetX check to the beginning of the loop.
* Calculate exact image size in GenImageFontAtlas
Calculate exact image size with a method based on total glyph width and glyph row count
Current method seemed a little bit overkill with square root, log and power functions and only approximates image size which can be wonky with some weird fonts like cursive fonts.
Proposed method calculates image size directly with a simpler method and results exact image size needed.
* Update rtext.c
* Update rtext.c
Changed do-while to while loop, and also added an extra step to calculate maximum glyph width and excluding it from image width for extra safety.
The `qoaplay_open()` function expects a `char *`, but we are passing in a `const char *`. While this works just fine, it does issue a compiler warning when strict:
```
src/raudio.c: In function ‘LoadMusicStream’:
src/raudio.c:1290:45: warning: passing argument 1 of ‘qoaplay_open’ discards ‘const’ qualifier from pointer target type [-Wdiscarded-qualifiers]
1290 | qoaplay_desc *ctxQoa = qoaplay_open(fileName);
| ^~~~~~~~
In file included from src/raudio.c:233:
src/external/qoaplay.c:86:34: note: expected ‘char *’ but argument is of type ‘const char *’
86 | qoaplay_desc *qoaplay_open(char *path)
| ~~~~~~^~~~
```
This change casts the argument to a `char *` to fix the warning.
I'm the only maintainer and have been working hard recently to get my library to be as nice to use as possible. I found out about this language list, and getting a new language on it didn't seem so bad. A bonus for my efforts.
