From: Ken Murchison (no email)
Date: Mon Sep 17 2007 - 10:15:25 EDT
Tomas Janousek wrote:
> Hello,
>
> On Mon, Sep 17, 2007 at 09:49:38AM -0400, Ken Murchison wrote:
>> I applied Simon's version to CVS, but then I realized that it causes
>> myfreestate() to crash on my Linux box, because the newstate->groups array
>> doesn't get entirely populated. The following patch fixes the problem of
>
> Oh, sorry for that.
>
>> having gaps in newstate->groups, but I'm not sure why we're skipping groups
>> anyways. Also, will getgrouplist() really return -1 for anything other than
>> the output array being too small (wondering if we need the 'do' loop)?
>
> Yes, it returns -1 also in case of an error. It's not in the manpage but I
> looked at the glibc sources (at the time when I wrote the patch, so I may be
> inaccurate).
>
>> --- auth_unix.c.~1.42.~ 2007-09-13 13:24:42.000000000 -0400
>> +++ auth_unix.c 2007-09-17 09:38:45.000000000 -0400
>> @@ -270,11 +270,14 @@
>> goto err;
>> }
>>
>> - newstate->group = (char **)xmalloc(newstate->ngroups * sizeof(char *));
>> - for (i = 0; i < newstate->ngroups; ++i ) {
>> + newstate->ngroups = 0;
>> + newstate->group = (char **)xmalloc(ngroups * sizeof(char *));
>
> ngroups needn't be the same as newstate->ngroups, I think.
Right. I changed the getgrouplist() calls to write to ngroups, rather
than newstate->ngroups, so that ngroups always holds the most recent
value, and newstate->ngroups holds the previous value.
Can you explain the use of:
if (pwd || groupids[ngroups] != gid)
-- Kenneth Murchison Systems Programmer Project Cyrus Developer/Maintainer Carnegie Mellon University
|
|
|