[Buildd-tools-devel] Bug#520713: Bug#520713: init script start hangs

[Roger Leigh]

On Sun, Mar 22, 2009 at 02:53:04PM +0100, Peter Palfrader wrote:
> On Sun, 22 Mar 2009, Peter Palfrader wrote:
> 
> Core was generated by `schroot --chroot=sid-alpha-sbuild-7855514e-e68e-40e3-8198-61366cd44074 --recove'.
> Program terminated with signal 6, Aborted.
> [New process 24735]
> #0  0x0000020000452d68 in raise () from /lib/libc.so.6.1
> (gdb) bt
> #0  0x0000020000452d68 in raise () from /lib/libc.so.6.1
> #1  0x0000020000454664 in abort () from /lib/libc.so.6.1
> #2  0x00000200004498ec in __assert_fail () from /lib/libc.so.6.1
> #3  0x00000200068628ec in __pthread_tpp_change_priority () from /lib/libpthread.so.0
> #4  0x0000020006858fc4 in pthread_mutex_lock () from /lib/libpthread.so.0
> #5  0x0000020000511124 in pthread_mutex_lock () from /lib/libc.so.6.1
> #6  0x0000000120078200 in chroot (this=0x12014dd68) at /usr/include/boost/detail/sp_counted_base_pt.hpp:76
> #7  0x0000000120088474 in sbuild::chroot_plain::clone (this=0x1201399b0) at ../../../sbuild/sbuild-chroot-plain.h:31

Thanks.  Now, this is interesting.

clone() creates a copy of a chroot object, returning is as a
std::tr1::shared_ptr<>.  This is part of libstdc++.

However frame #6 shows it in inline code in one of the boost headers.

Possibly a mismatch between the C++ TR1 and Boost headers?

Possibly a locking bug in libstdc++ or libc6 on Alpha?

This isn't, AFAICT now, a bug in schroot.  It's actually in the
shared_ptr locking code, which makes sense.  The locks would need
to be thread safe.

Where to go from here?
Firstly, are we missing any compiler/linker options to be thread-safe?
While we are linked against pthreads, do we need to initialise them
explicitly in order to use them safely?


Regards,
Roger

-- 
  .''`.  Roger Leigh
 : :' :  Debian GNU/Linux             http://people.debian.org/~rleigh/
 `. `'   Printing on GNU/Linux?       http://gutenprint.sourceforge.net/
   `-    GPG Public Key: 0x25BFB848   Please GPG sign your mail.