Re: Ugh... regexp.c problems now...

Alex Manoussakis <[email protected]> Sat, 7 Jan 2012 08:24:26 -0800
Newsgroups gmane.comp.bug-tracking.gnats.general
Message-ID <[email protected]>
One thing I had to do in getdate.y is this:

diff -u -p -r1.8 getdate.y
--- getdate.y	30 Aug 2007 05:53:05 -0000	1.8
+++ getdate.y	7 Jan 2012 16:22:02 -0000
@@ -44,8 +44,8 @@

  #include <ctype.h>

-#if HAVE_STDLIB_H
-# include <stdlib.h> /* for `free'; used by Bison 1.27 */
+#if STDC_HEADERS || HAVE_STDLIB_H
+# include <stdlib.h> /* For `abort'. Also for `free'; used by Bison 1.27 
*/
  #endif

  #if STDC_HEADERS || (! defined isascii && ! HAVE_ISASCII)


anyway, for external code like regex.c we could disable warnings,
and only have them on for stuff we "own".

Thanks,
Alex

On Sat, 7 Jan 2012, Chad Walstrom wrote:

> I've been going down one rabbit hole and into another.  One comment I
> saw on the web stated outright that flex (flex.sourceforge.com) was
> simply couldn't be compiled with "-Wall".  I've come up with a possible
> solution with flex by invoking a couple of options:
>
> %option never-interactive
> %option noinput
>
> The first one seems reasonable, but I'm not sure about the second.  It
> does clear the compiler errors and allows us to continue on. That simply
> uncovered a number of bugs with getdate.y.  After wrestling with that
> for a while, I opted to try out gnulib's getdate implementation. It
> allowed me to get a little further, and BOOM, regexp.c blew up
> fantastically.  Well, there is a regexp module in gnulib as well...
>
> But it's way to late tonight to go down that route.
>
> Chad
>