Re: some fixes to musl

[Vasiliy Kulikov <segoon@openwall.com>]

On Fri, Jul 22, 2011 at 00:30 -0400, Rich Felker wrote:
> On Thu, Jul 21, 2011 at 09:57:39PM -0400, Rich Felker wrote:
> > > forkpty():
> > > - It should be guaranteed that master fd is closed, tty is setup, slave
> > >   fd is dup'ed to 1,2,3.  The latter can be broken by setting small
> > >   rlimit.  setsid() is checked for company :)  I think the only way to
> > >   handle the failure is _exit().  While it may be not the best choise,
> > >   however, continuing the work with half dropped privileges is more
> > >   dangerous.
> > > 
> > > openpty():
> > > - close() shouldn't change errno updated by failed ioctl()/open().
> > > - I suppose the last calls to tcsetattr() and ioctl() may fail too.
> > 
> > Going to try to find a good solution for these...
> 
> I believe I've fixed forkpty's issue with fd exhaustion. Please tell
> me if anything seems wrong.

Looks like it fixes the problem with *expected* dup2() failures.
However, I'm still worried about theoretical (*m) leaking...

-- 
Vasiliy