On 11/23/2016 08:17 AM, Markus Armbruster wrote:

> 
> The first two patches are bug fixes, and as such they should be
> considered for 2.8.
> 
> This patch doesn't fix anything, and it might conceivably break
> something.  Too late for 2.8.

Ah, but it DOES fix check-qjson on Mac OS.

As mentioned to Paolo, I'm splitting this into two parts for the v2
series (the first part to fix testsuite failures on Mac OS which is
still 2.8 material, the second to rip out %I64d which becomes more of
2.9 material).

>> My other argument is that I _do_ intend to rip out ALL of the dynamic
>> JSON support, at which point we no longer have %d, let along %lld.
>> Until you see that followup series and decide whether it was too
>> invasive for 2.9, it's hard to say that we are throwing out anything
>> useful in this short-term fix for 2.8.  So I guess that gives me a
>> reason to hurry up and finish my work on that series to post it today
>> before I take a long holiday weekend.
> 
> If we rip out _jsonf() in 2.9, then ripping out currently unused parts
> of it in 2.8 during hard freeze is needless churn at a rather
> inconvenient time.
> 
> If we decice not to rip it out, it may well have to be reverted.
> 
> I don't think there's a need to hurry, as this patch isn't appropriate
> for 2.8 anyway, so there's no reason to quickly decide what to do with
> the followup series now.

Fair enough.

v2 coming soon.

-- 
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org