Hajimu UMEMOTO wrote:
> Hi,
>
>>>>>> On Thu, 25 Oct 2007 23:03:35 -0400
>>>>>> Ken Murchison <murch@xxxxxxxxxxxxxx> said:
>
> murch> John Capo wrote:
>> On Thu, October 25, 2007 21:10, John Capo wrote:
>>> Quoting Ken Murchison (murch@xxxxxxxxxxxxxx):
>>>
>>>> Simon Matter wrote:
>>>>
>>>>>> On the Linux box, all fresh compilations aside from the sasl 2.1.15 binaries:
>>>>>>
>>>>> I once posted to the list that 2.3.9 needs at least cyrus-sasl-2.1.19. As a package
>>>>> maintainer I know that :)
>>>> Did you ever figure out why? I'm not surprised that code in Cyrus somehow depends on
>>>> a change in SASL, but I can't seem to find anything in the CVS logs or diffs that
>>>> would be the cause.
>>> This is what I had to do for cmd_login to work in 2.3.9.
>>>
>>>
>>> /* authstate already created by mysasl_proxy_policy() */
>>> /* Not when using login and allowplaintext. imapd_authstate is NULL TM Login fix */
>>> if (imapd_authstate == NULL)
>>> imapd_authstate = auth_newstate(imapd_userid);
>>>
>>> But 2.3.10 cores :-(
>> Its coring in getgrouplist() probably because the 3rd argyument is NULL.
>>
>> /* get number of groups user is member of into ngroups */
>> getgrouplist(identifier, gid, NULL, &ngroups);
>>
>> BSD man page does not indicate that NULL args are OK.
>>
>> int
>> getgrouplist(const char *name, int basegid, int *groups, int *ngroups);
>>
>> The resulting group list is returned in the integer array pointed to by
>> groups. The caller specifies the size of the groups array in the integer
>> pointed to by ngroups; the actual number of groups found is returned in
>> ngroups.
>
>
> murch> See if this fixes the getgrouplist() problem:
>
> murch> --- auth_unix.c.~1.46.~ 2007-09-27 16:02:45.000000000 -0400
> murch> +++ auth_unix.c 2007-10-25 23:02:15.000000000 -0400
> murch> @@ -225,7 +225,7 @@
> murch> struct group *grp;
> murch> #ifdef HAVE_GETGROUPLIST
> murch> gid_t gid, *groupids = NULL;
> murch> - int ret, ngroups = 0;
> murch> + int ret, ngroups = 10;
> murch> #else
> murch> char **mem;
> murch> #endif
> murch> @@ -248,10 +248,7 @@
> murch> #ifdef HAVE_GETGROUPLIST
> murch> gid = pwd ? pwd->pw_gid : (gid_t) -1;
>
> murch> - /* get number of groups user is member of into ngroups */
> murch> - getgrouplist(identifier, gid, NULL, &ngroups);
> murch> -
> murch> - /* get the actual group ids */
> murch> + /* get the group ids */
> murch> do {
> murch> groupids = (gid_t *)xrealloc((gid_t *)groupids,
> murch> ngroups * sizeof(gid_t));
>
> The NGROUPS is defined in sys/param.h on BSDs as:
>
> #define NGROUPS NGROUPS_MAX /* max number groups */
>
> The NGROUPS_MAX is defined in sys/syslimits.h as:
>
> #define NGROUPS_MAX 16 /* max supplemental group id's */
>
> and, sys/syslimits.h is included from sys/param.h. So, you should use
> this instead of a magic number, IMHO.
From limits.h on my Fedora box:
#define NGROUPS_MAX 65536
It seems like a waste of memory to use NGROUPS_MAX as the default size
on this platform.
--
Kenneth Murchison
Systems Programmer
Project Cyrus Developer/Maintainer
Carnegie Mellon University
----
Cyrus Home Page: http://cyrusimap.web.cmu.edu/
Cyrus Wiki/FAQ: http://cyrusimap.web.cmu.edu/twiki
List Archives/Info: http://asg.web.cmu.edu/cyrus/mailing-list.html
