dbus crashing with no environment

[Bastien Nocera]

On Tue, 2008-11-25 at 15:59 +0000, Daniel P. Berrange wrote:
> On Tue, Nov 25, 2008 at 03:18:40PM +0000, Bastien Nocera wrote:
> > Heya,
> > 
> > I was trying to see why fprintd was crashing when using the service
> > activation, and Richard mentioned that I should try clearing the
> > environment as that's what the activation helper does.
> > 
> > I don't actually need any environment in fprintd, so went ahead with it,
> > and ended up with crashes in dbus itself. fprintd uses dbus-glib, but I
> > was able to reproduce with the attached DBus-only test case.
> 
> This does not appear to be neccessarily dbus's  fault - change your
> demo program to just be 
> 
>   #include <stdlib.h>
> 
>   int main (int argc, char **argv)
>   {
>         clearenv ();
>         unsetenv("FOO");
>         return 0;
>   }
> 
> 
> And it'll also crash & burn the same way:

Duh. My mistake for not going deeper into it. Would you mind filing a
bug against glibc, or shall I do it?

> ==14998== Invalid read of size 4
> ==14998==    at 0xB3A53F: unsetenv (setenv.c:295)
> ==14998==    by 0x4008326: unsetenv (mc_replace_strmem.c:760)
> ==14998==    by 0x8048415: main (env.c:6)
> ==14998==  Address 0x0 is not stack'd, malloc'd or (recently) free'd
> 
> So its arguably a bug in libc's unsetenv(), but we could guard against 
> it in DBus by making the dbus_setenv do
> 
>    if (getenv("FOO")) unsetenv("FOO");
> 
> Instead of calling unsetenv() unconditionally.

Probably a good idea.