> I set the memory options back, removed %option noyywrap, and added
> #undef yypwrap to the header section. This appears to work in Solaris
> and in flex on my linux system. Here's the patch...
Did you check these in? They didn't appear to be in the CVS version.
Anyway, I added back the memory lines. Instead of #undef-ing yywrap, I
put in a #define YY_SKIP_WWRAP and a dummy #define yywrap() 1, which
is effectively what the flex %option noyywrap does, so we should be
all good.
> If it's going to turn out the same file, then obviously there's no
> point in running it. I wonder though if flex on Solaris will output
> the same file as flex on SunOS, or flex on linux, etc... Does flex
> have any built-in system-dependent localizations? I just don't know.
Yeah, Dan's argument of signed-char's and other things seems
reasonable. I'll put a test into the Makefile to use flex/lex if it
exists (I'm guessing that autoconf will pick flex over lex, but I need
to check on that.), else use the pre-lexed version.
I've also taken out the case-insensitive flag to lex and fixed
dtimep.lex accordingly.
Shantonu
