regcomp(3) incorrectly handles chars with 8th bit set in format string

[Vlad Harchev]

 Hi, 

 Sorry, I can't provide you with exact version number of glibc (I'm writing
from other place, not from where I got a bug) - but it's about glibc shipped
with non-upgraded (vanilla) RedHat 7.1 on x86.

 Bugreport itself:

 regcomp(3) incorrectly handles characters with 8th bit set in format string.

 This leads to crash (due to a bug in wctomb - I've reported it to you in
other message).
 Here is a backtrace (sorry for extra quoting):

> > (no debugging symbols found)...0x40864519 in __wait4 ()
> >    from /lib/i686/libc.so.6
> > #0  0x40864519 in __wait4 () from /lib/i686/libc.so.6
> > #1  0x408d49e4 in __DTOR_END__ () from /lib/i686/libc.so.6
> > #2  0x404a3262 in gnome_segv_handle (signum=11) at gnome-init.c:659
> > #3  <signal handler called>
> > #4  0x40956f73 in gconv () from /usr/lib/gconv/KOI8-R.so
> > #5  0x4083688a in __wcrtomb (s=0xbffff590 "\210\002", wc=-15,
> > ps=0x408d8aa0)
> >     at wcrtomb.c:85
> > #6  0x407da8cf in wctomb (s=0xbffff590 "\210\002", wchar=-15) at
> > wctomb.c:53
> > #7  0x40875bd8 in __re_compile_fastmap (bufp=0xbffff630) at regex.c:4567
> > #8  0x40879c24 in __regcomp (preg=0xbffff630,
> >     pattern=0x8201720
> >
"^(Янв|Фев|Мар|Апр|Май|Июн|Июл|Авг|Сен|Окт|Ноя|Дек)-([0-9][0-9]?)-([0-9][0-9]?)$",
> > cflags=3) at regex.c:7745
 
 I assume that macro regcomp expands to __regcomp because frame #9 is about
user function.

 It's under ru_RU.KOI8-R locale (but I guess under any locale it will be the
same).

 Solution:
 Instead of
        (wchar_t)some_character
 use
        (wchar_t)(unsigned char)some_character
 This way
        wctomb will get a positive argument.
 But wctomb also needs fixing.

 Thanks for developing glibc!

 Best regards,
  -Vlad