[EMAIL PROTECTED] - Sun Feb 27 03:56:20 2005]: > On Sun, Feb 27, 2005 at 02:13:59AM +0100 I heard the voice of > Richard Levitte via RT, and lo! it spake thus: > > > > In Imakefile, check what happens when I18N is defined. It's > > undefined by default. > > > > Actually, I think the default is silly, so I'd like to ask the > > general public if there would be any negative effects if I18N was > > defined by default. > > The FreeBSD port has built with I18N turned on for 2 years and change.
OK. I suspect it might have to do with the input functions in parse.c returning int, and getting the characters sign extended to int (and thus become negative values). Trouble is, in lex.l, there's a conditional definition of YY_INPUT, and in the I18N case, it looks pretty sane, even when given "negative" characters. This is, however, only when using flex if I understand correctly. Also, m4 might be involved in the equation. So, the question is, what lex implementation do you use, and is m4 involved in reading the configuration file? /Richard
