[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

debugging broken with patched libc [was: Re: POSIX RW locks (pthread_rwlock_wrlock) issues on amd64/sarge 64bit.

On Thu, 16 Jun 2005, Karel Gardas wrote:

On Sat, 11 Jun 2005, Kurt Roeckx wrote:

On Fri, Jun 10, 2005 at 07:47:46PM -0300, Javier Kohen wrote:
Hello Karel,

El vie, 10-06-2005 a las 22:04 +0200, Karel Gardas escribió:

I do have amd64 box with relased sarge/amd64 installed. I'm using
gcc-3.4.x from debian package and I have multi-threading application which
strangely blocks on the call to pthread_rwlock_wrlock function. The
problem is, I'm 100% sure the lock is not hold by any reader, so wrlock
should succeed and continue w/o blocking. I'm also sure that this software
shouldn't be buggy in this domain, since it happily runs in 32bit chroot
of sarge compiled by the same compiler (just 32bit inside chroot) and also
it runs well on 64bit solaris10 compiled by sunstudio10 compiler into
64bit executables.

There is a TV app whose name I don't remember that I seem to recall was
supposed to suffer the same problem on AMD64. There was a patch floating
around that changed the rwlock calls to use a different locking
mechanism, but I don't know if anybody was working on fixing the real

Could someone please file a bug against the libc6 package about
this if it doesn't exist yet?

I've gone ahead and found it already reported:


just to followup with this thread. I've used the patch pointed out by the bugreport above, rebuilt glibc, installed just libpthread.so by doing manual cp and found that it really solves the issue with RW locks. The problem now is that debugging of multi-threaded applications is broken, i.e. `info threads' prints nothing and `thread <num>' is not working.

Does anybody here have an idea what's going wrong with the debugger or "hacked" libc?

Karel Gardas                  kgardas@objectsecurity.com
ObjectSecurity Ltd.           http://www.objectsecurity.com

Reply to: