In file included from .sconf_temp/conftest_12.cpp:3: /usr/include/jpeglib.h:792:3: error: unknown type name 'size_t' size_t free_in_buffer; /* # of byte spaces remaining in buffer */ ^ [...] /usr/include/jpeglib.h:1000:57: error: unknown type name 'FILE' EXTERN(void) jpeg_stdio_dest JPP((j_compress_ptr cinfo, FILE * outfile)); ^ /usr/include/jpeglib.h:895:22: note: expanded from macro 'JPP' ^
When not using rlbabel, we always measure characters in terms of MOJI_SIZE instead of the actual font's metrics. Notably, several Japanese fonts on Linux have completely busted minx/advance metrics for western characters, and should not be consulted. The current text breaking rule appears to be: break a line if I can't fit another fullwidth character unless the next character is a space. This also removes a block which was inserting a fullwidth space at the beginning of Yumemi's lines. (This may be controlled by a Gameexe.ini key though; when I have more time, I should double check that this doesn't introduce a regression in Japanese planetarian.) This fixes text output on at least the first 30 screens of text in the English planetarian. (This also fixes Japanese Clannad, which was putting one less character on lines than it should have been.)
For now, I'm just copying the implementations. I haven't seen any semantic differences yet.
Either Haeleth was wrong originally about the size of the strK bank, or RealLive changed the maximum size. Either way, make the strK baking a vector, and have it autosize itself.
…king. It looks like RealLiveMax is just like Kinetic: the actual instructions are implemented the same, but have different opcode numbers.
This reverts commit 740985e. This breaks the title screen on CLANNAD when there are several lights.
I've hacked something equivalent to this into rlvm's code going on four times. Time to polish it and make it a real flag.
In the Kud Wafter ED, we start with a zoom out which had some blockyness from being too zoomed in.