Dear Maintainer, I can confirm that the archived bug #773503 is not fixed.
libvirtd still segfaults on live migration.
Last message in #773503 may give some clue.

Thank you!

Did you test the patch in this bug report successfully?
 -- Guido

Unfortunately, no. But I may get a window of opportunity in the next couple of days.

The snapshot at:

http://honk.dyn.sigxcpu.org/projects/libvirt/snapshots/

is built for jessie and has these changes applied.
Cheers,
 -- Guido

I wanted to test jessie version, but it fails to build (even without the patch):

FAIL: virsh-uriprecedence
=========================

  1) User config file                                             ... SKIP
     case skipped: must not be run as root
  1) LIBVIRT_DEFAULT_URI                                          ... FAILED
  2) VIRSH_DEFAULT_CONNECT_URI                                    ... FAILED
  3) Parameter                                                    ... FAILED

Any hints about where to look?

Guido, thanks, but your link became unaccessible.

At libxml2 it has a bug in jessie. Use the debian/jessie branch from git
or my pre built snapshorts.

Please try again:

http://honk.sigxcpu.org/projects/libvirt/snapshots/

It surely works from here. Alternatively you can try:

http://honk.dyn.sigxcpu.org/projects/libvirt/snapshots/

which is one hop less and also reachable via ipv6.

Cheers,
 -- Guido

OK, tried to migrate a VM to the host with libvirt version from your snapshot. It did not segfault, so the patch appears to be working.

But migration still fails (version difference?)

sudo virsh migrate --live --persistent --copy-storage-all --p2p --tunneled vm1 qemu+ssh://root@vhost2/system --verbose
Migration: [ 85 %]error: operation failed: migration job: unexpectedly failed

It seem to successfully copy block device, define VM, start it in paused state, but in the end this:

in qemu log:

Receiving block device images
#( there was a line 'Completed 99 %' here)
2015-06-14 11:47:07.566+0000: shutting down
2015-06-14T11:47:07.616812Z qemu-system-x86_64: load of migration failed: Input/output error

in journal:

libvirtd[55004]: error from service: TerminateMachine: No machine 'qemu-vm1' known

Thanks for testing!

I have currently no idea what causes this. I will have to set up a test
system to check but this will take some time. Do you have a chance to
check if it works with the version in sid?

This one is ugly but harmless.
Cheers,
 -- Guido

I could test migration from Jessie to Stretch host (I am low on extra machines now). If that would be a valid test.

I'll try to set up to Jessie systems here. Will take some time though,
at least we don't crash the daemon anymore with that path
applied. Thanks a lot for testing!
Cheers,
 -- Guido

I have two stretch machines and --live migration with active --copy-storage-all won't work. I don't know if this refers to this (old) but not closed bug. There's nothing segfaulting when I try to migrate, but the migration won't start and got stuck on both machines. When starting a live migration, f.i. this way

virsh migrate --live --desturi qemu+tls://host2/system --copy-storage-all --verbose --persistent --undefinesource test2

the command doesn't come back. On the host2, the machine got created in "pause" mode. This process never ends and can only be canceld by restarting libvirtd on host2.

Just to get you informed:

I found adding --tunnelled and --p2p like this

virsh migrate --live --desturi qemu+ssh://host2/system --copy-storage-all --p2p --tunnelled --verbose --persistent --undefinesource test2

does the job for me and live migration to the second job starts and succeed...

Greetings

Micha