linuxthreads bug in 2.2.4 under ppc linux

[Kevin B. Hendricks]

Hi Wolfram,

> This sounds suspiciously like the problem I analyzed for months in
> summer.  Are you mixing fork() and threads?  That's how I stumbled
> over the problem on Intel.  Basically, with glibc-2.2.4 a 'phantom'
> thread can be created and hold locks, with disastrous consequences.
> The fix went in only shortly after 2.2.4 was released.

Yes, in fact this code is playing around with fork() and threads.  The 
thread in question has just forked a child and moved on to check for the 
reap and grab its child's status when the problem hits.

> Please try the current glibc-2.2 branch from CVS, or for example the
> Debian-packaged version of glibc-2.2.4.

Can you pont me to a specific patch that I can check for?  I think Franz 
said the build tree he based his rpm on was very very recent.

Kevin