Re: Fwd: Re: [PATCH 21.5] Fix signs in time zone string.

>>>>>"AA" == Adrian Aichner <[email protected]> writes:
AA> How could XEmacs and GNU Emacs both be wrong until now?

XEmacs wasn't wrong for long, this caused the problem:

changeset:   5762:427a72c6ee17
user:        Stephen J. Turnbull <[email protected]>
date:        Sun Sep 15 23:47:37 2013 +0900
summary:     Eliminate several compiler (clang) warnings.

diff --git a/src/editfns.c b/src/editfns.c
--- a/src/editfns.c
+++ b/src/editfns.c
@@ -1190,7 +1190,9 @@
       else if (FIXNUMP (zone))
        {
          int abszone = abs (XFIXNUM (zone));
-         sprintf (tzbuf, "XXX%s%d:%02d:%02d", "-" + (XFIXNUM (zone) < 0),
+         /* #### I have no idea what this conforms to,
+            but the compiler has stopped whining. */
+         sprintf (tzbuf, "XXX%s%d:%02d:%02d", (XFIXNUM (zone) < 0) ? "-" : "+",
                   abszone / (60*60), (abszone/60) % 60, abszone % 60);
          tzstring = tzbuf;
        }

AA> http://git.savannah.gnu.org/cgit/emacs.git/tree/src/editfns.c#n1937
AA> suggests that they are not including a + character for positive offsets,
AA> which surprised me.

According to `man tzset', the + character is optional.  I prefer to make
it explicit to avoid confusion.

AA> But zone offsets > 0 do not include the - sign, which is the same logic
AA> XEmacs uses before your patch.

No, it is not.  GNU Emacs is correct, but it's implementation is a
rather hacky array access:  &"-"[XINT (zone) < 0]

A negative zone offset yields &"-"[1], which skips the minus and yields
the empty string; a zero or positive zone offset yields the minus via
&"-"[0].

AA> Also, why are you removing the REST argument in passing without mention?

That was a misunderstanding on my part, sorry for that.

-- 
Marcus