2.0.5a core dump in setup_groups()

[Jeremy Allison]

Ken Weaverling wrote:
> 
> A few weeks ago, I posted here that 2.0.4b dumped core on me in getpwent
> calls in certain rare, but reproducable cases. getpwent was triggering a
> realloc() which cause the crash.
> 
> Well, I noticed a lot of work in that area in 2.0.5a but guess what, it
> now dumps core in a different area, relating to getgrent and initgroups.
> Why, I have no idea.  Curiously though, it's using the same reproducable
> set of steps.  It seems to work all but a few others.
> 
> This is on an m88K DG/UX 4.11MU05 system.
> 

I've just worked with someone else who has a reproducible
crash bug with getpwent when a user is in more than 8
groups on a DG/UX machine. Set the groups to 8 and
the problem goes away. It's not an smbd problem (I tested
it on IRIX :-).

It's starting to look like DG/UX has some C library bugs
relating to memory overwrites with limits like MAX_GROUPS.

Could this possibly be a similar problem here ?

Jeremy.

-- 
--------------------------------------------------------
Buying an operating system without source is like buying
a self-assembly Space Shuttle with no instructions.
--------------------------------------------------------