Hi Gerel. You may have a point but I can't see it. See the output I get when changing pdf_char_t to 'char' with the -funsigned-char flag and without it (on intel).
[...] I don't consider either one as a 'soulution' per se, I mean, if we start using that flag, our code will be full of "(signed char*)". Am I wrong ? It would not be needed because we are not going to use pdf_char_t values as signed ones. With -funsigned-char the system functions expecting char* would expect 'unsigned char*', so we could pass pdf_char_t* values (that are currently unsigned) to these functions without having to do an explicit cast to avoid the warning launched by -Wall. Anyway, this is a minor issue. Programs that depend on -funsigned-char are less portable than programs that dont depend on it, so at this point I would say to go ahead and continue using the explicit casts when calling system functions. If you agree please close the task with a DISCARDED state. We keep casting like god orders. We have no gods in the church of Emacs ;)
