"bug in glibc"

[Justin Cormack]

On Sun, Mar 31, 2013 at 4:12 PM, Igor Sysoev <igor at sysoev.ru> wrote:

> On Mar 31, 2013, at 14:33 , Justin Cormack wrote:
>
> > There is a note in src/os/unix/ngx_user.c about a bug in glibc for
> crypt_r:
> >
> >     /* work around the glibc bug */
> >     cd.current_salt[0] = ~salt[0];
> >
> >     value = crypt_r((char *) key, (char *) salt, &cd);
> >
> > I was wondering if anyone knew what the bug was, as I am running on a
> platform (Musl libc) that has got NGX_HAVE_GNU_CRYPT_R but has a different
> implementation, in particular has no current_salt field in struct
> crypt_data (and indeed the man page says you should treat it as opaque
> except for the initialized field).
> >
> > I was wondering exactly what the bug was as then I could write a test
> for it rather than always including this code; I have not been able to find
> it in the glibc bug tracker though.
>
>
> 2002-10-29  Daniel Jacobowitz  <drow at mvista.com>
>
>         * crypt/crypt_util.c (__init_des_r): Initialize current_salt
>         and current_saltbits.
>
>
> https://groups.google.com/forum/?fromgroups=#!topic/linux.debian.maint.glibc/Q88bwAp222w
>
>
Ok I just checked and this bug was fixed in glibc-2.3.2, which was released
in March 2003... Any chance of removing the workaround as it is relying on
fields that may not exist and are implementation internal?

Justin
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.nginx.org/pipermail/nginx/attachments/20130331/6a463ad4/attachment.html>