Compatibility is dead

[Daniel van Vugt]

As is often the case when I make too many public comments, they quickly 
become outdated. I've been using gcc-4.8 on raring since yesterday :)


On 16/07/13 13:41, Thomas Voß wrote:
> Hey Daniel,
>
> the one concern I have with maintaining gcc 4.7 compatibility are the
> numerous bugs that trigger internal compiler errors quite frequently.
> Those issues are mostly gone with 4.8 and I would want to avoid us
> working around compiler bugs (as opposed to missing C++11 features).
>
> On Tue, Jul 16, 2013 at 3:39 AM, Daniel van Vugt
> <daniel.van.vugt at canonical.com> wrote:
>> My main desktop is indeed raring. However it is significantly outnumbered by
>> the number of saucy machines I have. I like raring because it is my fastest
>> build machine by far. So keeps me more productive.
>>
>
> Fair point, why not setup a saucy chroot on the machine? Works
> perfectly fine for me. Or a gcc 4.8 from the toolchain ppa's on
> launchpad and leveraging the compute power with the help of distcc?
>
>> I'm not going to try to justify compiler compatibility any more other than
>> saying it's a good thing for any project.
>
> While I do agree in general, I'm opposed to investing too much time
> working around gcc 4.7 bugs.
>
> Thanks,
>
>    Thomas
>
>>
>> The gmock issues just look like CMake magic. So fixable, but not a priority.
>>
>>
>
>>
>> On 16/07/13 01:37, Oliver Ries wrote:
>>>
>>> Hi Daniel,
>>>
>>> thanks for pointing this out. Can you help me understand the impact a
>>> bit better?
>>>
>>> On Sun, Jul 14, 2013 at 9:31 PM, Daniel van Vugt
>>> <daniel.van.vugt at canonical.com <mailto:daniel.van.vugt at canonical.com>>
>>>
>>> wrote:
>>>
>>>      It finally happened. lp:mir now depends on C++11 features and
>>>      external packages (google-mock) too heavily to be buildable on
>>>      raring any more.
>>>
>>>      Essentially the main problem is that raring uses gcc-4.7 and Mir's
>>>      code depends on more complete C++11 features that only work in
>>>      gcc-4.8. The other issue is that Mir now depends on saucy-specific
>>>      hacks in the google-mock package, which can't be found elsewhere.
>>>
>>>      This is a little disappointing. If anyone would like to take on the
>>>      project of making the Mir source more backward compatible, then feel
>>>      free. But maintaining compatibility with raring has now taken up too
>>>      much of my time...
>>>
>>>
>>> Looking at your initial mail, the main motivator for offering to support
>>> Raring was that your own development environment is not on saucy yet
>>> (though it should;). From this mail I take that gcc 4.8 and gmock test
>>> are not available on R but became a requirement.
>>>
>>> Is there any other benefit that we can think of that is currently being
>>> impacted by this incompatibility with R?
>>>
>>> thx
>>> Olli
>>
>>
>> --
>> Mir-devel mailing list
>> Mir-devel at lists.ubuntu.com
>> Modify settings or unsubscribe at:
>> https://lists.ubuntu.com/mailman/listinfo/mir-devel