On 21 Jan 2014, at 11:51, Simon Hausmann <simon.hausm...@digia.com> wrote:

> On Tuesday 21. January 2014 10.23.22 Sorvig Morten wrote:
>> On 21 Jan 2014, at 09:32, Simon Hausmann <simon.hausm...@digia.com> wrote:
>>> On Monday 20. January 2014 20.21.14 deDietrich Gabriel wrote:
>>>> On Jan 20, 2014, at 7:55 PM, Allan Sandfeld Jensen <k...@carewolf.com> 
> wrote:
>>>>> On Monday 20 January 2014, Thiago Macieira wrote:
>>>>>> On segunda-feira, 20 de janeiro de 2014 17:36:26, Kurt Pattyn wrote:
>>>>>>> The CI system is still building for OSX 10.6.
>>>>>>> Given the fact that OSX is at version 10.9 now, shouldn’t the build
>>>>>>> for
>>>>>>> 10.6 be removed, and ideally replaced with a build for OSX 10.9?
>>>>>> 
>>>>>> Only if we decide to stop supporting 10.6 entirely. So the question is:
>>>>>> do
>>>>>> we drop it?
>>>>>> 
>>>>>> Mac devs, what say you?
>>>>> 
>>>>> Could we drop support for build on 10.6, but still support it as runtime
>>>>> platform? That seems to be how Apple prefers to support older versions.
>>>>> The
>>>>> question is of course if out CI system would be able to handle building
>>>>> on
>>>>> 10.7 but running  on 10.6
>>>> 
>>>> IIRC, this has been the official statement since Qt 5.0. And yet, for
>>>> some
>>>> reason, the CI is treating 10.6 as any other platform.
>>>> 
>>>> If you do the math from the data available here
>>>> http://www.netmarketshare.com/operating-system-market-share.aspx?qprid=10
>>>> &q
>>>> pcustomd=0 (that’s December 2013), 10.6 accounts for slightly less than
>>>> 20%
>>>> of all the OS X versions. Let’s suppose those numbers reflect the
>>>> reality.
>>>> 
>>>> Is 20% a lot? I don’t know. Is that 20% the same proportion for Qt 5
>>>> end-users? Or Qt 5 developers making a life out of it? Absolutely no
>>>> idea.
>>>> Do I see that many bugs reported on 10.6? No, not at all.
>>>> 
>>>> The truth is, market share doesn’t mean anything. Point in case:
>>>> According
>>>> to the link above, OS X is less than 8% of the total market share. Should
>>>> we then drop the Mac port completely?
>>>> 
>>>> For all I know, that twenty-something percent of Mac users running 10.6
>>>> or
>>>> earlier are all grandma and grandpa who only use Safari, Skype, and maybe
>>>> some spreadsheet software for tax returns. For all I know, none of them
>>>> use
>>>> any Qt 5 based software. And why would they? They haven’t updated their
>>>> system in years… But maybe I’m wrong, and I’d like to be proven so.
>>>> 
>>>> So, can someone tell me what that 20% really means for Qt 5 and its
>>>> developers? This is what it means for me.
>>>> 
>>>> We are not that many working on the Mac port. I can think about 6-8
>>>> people,
>>>> including me, and AFAIK none of us works 100% on the Mac port (I’d say
>>>> the
>>>> average is below 50%, so that’s 3-4 full-time people maximum). And the
>>>> widgets Mac style code is a mess because we still support 10.6. And we
>>>> can’t still use ARC because of 10.6 and some old Xcode version. And have
>>>> you seen the CoreWLAN bear management plugin and how we support 10.6?
>>>> Fullscreen mode hacks? Thank 10.6. Building WebKit and C++ 11, anyone?
>>>> 
>>>> Don’t give me “But 20% market share” or “The XP of Apple” when the debate
>>>> about ending support for 10.6 comes. Give me facts. Give me numbers that
>>>> concern Qt 5. Give me reasons why we should keep parts of Qt in such
>>>> unsatisfactory state.
>>>> 
>>>> I only work on my little things, and 10.6 is a burden for me. So, you who
>>>> work out there, that see people using Qt 5 apps, tell me, is it worth it?
>>> 
>>> I wholeheartedly agree with Gabriel.
>>> 
>>> We all are contributing to the Qt project - as opposed to private forks of
>>> Qt - because we want to grow the overall success of Qt, we want it to
>>> become even more popular among software developers. However we do have
>>> limited resources, so when we decide to spend time on something, it is
>>> very important to ask ourselves: How does for example working on 10.6
>>> contribute to the popularity of Qt compared to making it kick-ass on more
>>> recent versions of Mac OS X?
>>> 
>>> In my opinion the answer is crystal clear: We should provide first class
>>> integration with the latest Mac OS X technologies / frameworks, we should
>>> make life easier for application developers. I doubt that we can grow Qt
>>> faster by looking to the past - the future of Qt is more tightly
>>> connected to staying relevant and up-to-date with what's going on in the
>>> rest of the software industry.
>>> 
>>> 
>>> I personally like the suggestion brought up elsewhere of keeping things as
>>> they are for Qt 5.3 and dropping 10.6 from the CI system and supported
>>> platforms for 5.3 - while simultaneously reviewing and approving patches
>>> by
>>> other members of the community that continue to have an interest in 10.6
>>> support.
>> 
>> I agree with many of these arguments, and I was in favor of setting the
>> minimum supported version to 10.7 back when we started Qt 5 development.
>> But we did make the decision to support 10.6. The implementation effort has
>> been made and that decision should be respected.
>> 
>> Obviously it’s not going to stand forever, especially when seeing the strong
>> opinions from the Qt on Mac developers. We are moving in the direction of
>> not supporting 10.6. The 5.3 binary packages will not support it. QtWebkit
>> lives its own life - if upstream does not support 10.6 then there is little
>> we can do.
>> 
>> When we drop support for 10.6 we are going to remove the 10.6 code and start
>> using ARC. I don’t think there’s much room for "community support" for 10.6
>> - if we have to keep the manual reference counting code paths we might as
>> well fully support it.
> 
> Excellent points, that makes the choice a binary one indeed.
> 
> In your opinion - as the lead Qt on Mac guy - what should be the last 
> released 
> version of Qt to support deployment to 10.6?

That depends on how much time we spend releasing Qt :) 

I realize that if I’m the only one who want’s to keep supporting 10.6 then 
that’s not going to work. The most important thing to me is to have a somewhat 
predictable deprecation plan. For example (and at the risk of making this 
example “the plan”):

5.3 - Remove support from binary packages.
5.4 - 10.6 support is deprecated.
5.5? - Remove support.

Now you could argue that “deployment only” is de facto “deprecated”, but I 
think we should explicitly state it. Also, some time need to pass between 
“deprecated” and code removal, we can’t deprecate in 5.4 and then remove the 
code in dev the day after the release.

This thread should then be titled “Deprecate Mac OS 10.6 Build?”. The arguments 
for are:
- Parts of the dev team do not want to maintain it
- We want to free up CI resources 
- Questionable install base size

Sending a loud and clear “deprecated” message could actually help clear up that 
last point.

Morten

 


_______________________________________________
Development mailing list
Development@qt-project.org
http://lists.qt-project.org/mailman/listinfo/development

Reply via email to