It was used for error reporting only.

It was used for error reporting only.

Change all virsh commands that invoke virDomain{Attach,Detach}Device()
to use virDomain{Attach,Detach}DeviceFlags() instead.

Add a "--persistent" flag to these virsh commands, allowing user to
specify that the domain persisted config be modified as well.

V2: Only invoke virDomain{Attach,Detach}DeviceFlags() if
"--persistent" flag is specified.  Otherwise invoke
virDomain{Attach,Detach}Device() to retain current behavior.

Implementation of domain{Attach,Detach}DeviceFlags handlers
in the drivers.

Server side dispatcher for Domain{Attach,Detach}DeviceFlags.

Implementation of Domain{Attach,Detach}DeviceFlags in remote driver.

Definition of wire protocol format for
virDomain{Attach,Detach}DeviceFlags.

Implementation of public API for virDomain{Attach,Detach}DeviceFlags.

V2: Don't break remote compatibility with older libvirtd

Definition of internal API for virDomain{Attach,Detach}DeviceFlags.

Definition of public API for virDomain{Attach,Detach}DeviceFlags.

V2: Adjust libvrt_public.syms to anticipated 0.7.7 release.

virDomain{Attach,Detach}Device is now only permitted on active
domains.  Explicitly state this restriction in the API
documentation.

V2: Only change doc, dropping the hunk that forced the restriction
    in libvirt frontend.

* .gnulib: This fixes a warning in test-gettimeofday.c,
seen via "make check".  Reported by Daniel Veillard.

When configured with --enable-gcc-warnings, it didn't even compile.
* src/util/storage_file.c: Include <assert.h>.
(absolutePathFromBaseFile): Assert that converting size_t to int is valid.
Reverse length/string args to match "%.*s".
Explicitly ignore the return value of virAsprintf.

* src/util/storage_file.c: Include "dirname.h".
(absolutePathFromBaseFile): Rewrite not to leak, and to require
fewer allocations.
* bootstrap (modules): Add dirname-lgpl.
* .gnulib: Update submodule to the latest.

* src/xen/sexpr.c (sexpr_string): Free sexpr buffer upon allocation
failure.

* src/conf/domain_conf.c (virDomainChrDefParseXML): Don't leak a
virDomainChrDef buffer upon "unknown target type for char device" error.

* src/util/util.c (virExecWithHook): Free argv_str string before
returning upon failure to allocate space for environment.

* src/util/cgroup.c (virCgroupDetectPlacement): Close the mapping
FILE* also upon error.

* src/util/cgroup.c (virCgroupCpuSetInherit) [HAVE_MNTENT_H]:
Don't leak CPU-set inheritance value strings.

This stops the text and pre-boxes from touching the right border.

Instead of refering to __virRaiseError(VIR_ERROR_NO_MEMORY).

Also cleanup indentation of domain XML examples.

* tools/virt-pki-validate.in: use an automake variable instead of a
  shell variable so the resulting shell actually work

* configure.ac docs/news.html.in libvirt.spec.in: version bump and doc
  updates
* po/*.po*: updated and regenerated the localizations

When restoring from a saved guest image, the XML would already
contain the PCI slot ID of the IDE controller & video card.
The attempt to explicitly reserve this upfront would thus fail
everytime.

* src/qemu/qemu_conf.c: Reserve IDE controller / video card
  slot at time of need, rather than upfront

Similar fix as previous one but for fork() usage when creating
a file or directory

* src/util/util.c: virLogLock() and virLogUnlock() around fork()
  in virFileCreate() and virDirCreateSimple()

Ad pointed out by Dan Berrange:
So if some thread in libvirtd is currently executing a logging call,
while another thread calls virExec(), that other thread no longer
exists in the child, but its lock is never released. So when the
child then does virLogReset() it deadlocks.

The only way I see to address this, is for the parent process to call
virLogLock(), immediately before fork(), and then virLogUnlock()
afterwards in both parent & child. This will ensure that no other
thread
can be holding the lock across fork().

* src/util/logging.[ch] src/libvirt_private.syms: export virLogLock() and
  virLogUnlock()
* src/util/util.c: lock just before forking and unlock just after - in
  both parent and child.

The original udev node device backend neglected to lock the driverState
struct containing the device list when adding and removing devices
* src/node_device/node_device_udev.c: add necessary locks in
  udevRemoveOneDevice() and udevAddOneDevice()

* src/conf/interface_conf.c (virInterfaceDefParseProtoIPv4): If
virXPathNodeSet returns -1, indicate failure by returning -1 right away.
(virInterfaceDefParseProtoIPv6): Likewise.

* src/conf/storage_conf.c (virStoragePoolSourceListNewSource):
Remove an unused (and leaked) allocation.

* src/xen/xs_internal.c (xenStoreDomainIntroduced): Don't use -1
as an allocation size upon xenStoreNumOfDomains failure.
(xenStoreDomainReleased): Likewise.

If the primary security driver (SELinux/AppArmour) was disabled
then the secondary QEMU DAC security driver was also disabled.
This is mistaken, because the latter must be active at all times

* src/qemu/qemu_driver.c: Ensure DAC driver is always active

* src/xen/xen_hypervisor.c: Remove all "domain == NULL" tests.
* src/xen/xen_hypervisor.h: Instead, use ATTRIBUTE_NONNULL to
mark each "domain" parameter as "known always to be non-NULL".