On Tue, May 31, 2016 at 09:43:12 -0400, John Ferlan wrote:
On 05/31/2016 07:56 AM, Peter Krempa wrote:
> On Fri, May 27, 2016 at 09:57:07 -0400, John Ferlan wrote:
>> Move the module from qemu_command.c to virjson.c and rename to
>> virJSONBuildObjectCommandline.
>>
>> Nothing in the API is "specific" to the qemuBuild* processing and
it'll
>> be useful to "share" with upcoming changes in the storage_backend
which
>> will build a secret object for the 'qemu-img' command.
>
> It's very specific to qemu. Either qemu-img or qemu itself it's still
> qemu.
>
>> Also virJSONValueObjectForeachKeyValue and virJSONValueGetArrayAsBitmap
>> can be static helpers.
>>
>> Signed-off-by: John Ferlan <jferlan(a)redhat.com>
>> ---
>> src/libvirt_private.syms | 3 +-
>> src/qemu/qemu_command.c | 124 +++-----------------------------------------
>> src/qemu/qemu_command.h | 4 --
>> src/util/virjson.c | 117 +++++++++++++++++++++++++++++++++++++++--
>> src/util/virjson.h | 12 ++---
>> tests/qemucommandutiltest.c | 7 +--
>> 6 files changed, 129 insertions(+), 138 deletions(-)
>
> I'm leaning towards rejecting to add this to the JSON helpers. Maybe we
> need a new file for qemu specific helper functions to share accross
> modules.
>
Understood - the problem I keep running up against is "how much" to
"pull in". Creating some "shared" code library wasn't considered
yet,
but I can certainly give that a shot (munging with Makefile's is not
something I particularly look forward to). It was certainly the easy way
out to add code to src/util.
src/util/virqemu.(ch) ? It just doesn't belong to util/virjson.h in any
way.