[Libreoffice] Mac OS 10.4 Support

[Norbert Thiebaud]

On Fri, Jul 22, 2011 at 8:57 AM, Christian Lohmaier
<lohmaier+libreoffice at googlemail.com> wrote:
> Hi Norbert, *,
>
> On Fri, Jul 22, 2011 at 3:47 PM, Norbert Thiebaud <nthiebaud at gmail.com> wrote:
>> [...]
>> Caveat: Like tml_ I don't consider myself a 'Mac' developer... just a
>> Dev that happen to own a Mac.
>>
>> That being said, Christian is the person that need convincing.
>
> Oh, I don't consider myself any more Mac developer than you - I happen
> to have a PPC with 10.4 on it, so I'm affected (and I definitely won't
> update it to 10.5)
>
> I happened to be quite involved of getting rid of external
> dependencies on Mac, to compile against the SDK, but that's about it.
> Apart from a few bugfixes here and there and making it possible to
> uses internal python, I haven't been involved in any actual porting
> efforts.
>
> I just find that "drop it because it is old" is not enough reason to
> drop 10.4 support.
>
> And raising it to 10.5 won't help anything if you want to use 10.7
> features. You then have to provide different codepaths as well, and
> need runtime detection for that unless you want to release separate
> builds for different versions of Mac OS X - so where's the point?
> Where's the benefit of lifting the baseline to 10.5?

The benefit is that a new developer can still build the product, even
if he has _only_ a brand new lion.

As you mentioned earlier, You worked hard so that the Mac build, build
out-of-the box, without 'external dep'.
Having to hunt for a 10.4 SDK, and having to do some ln magic and
other trick to beat your box into submission is not going in that
direction.

Regarding different code path:

Yes, I think that it is eventually inevitable. I understand you
wanting to keep the product build-able on 10.4, but why does we have
to live with the lowest common denominator and not use a 'minimum
version' build flag. (I did hear the argument that if we do that, then
some mac dev can introduce, unknowingly, 10.4 incompatibility without
the proper flag isolation... but that is the very situation that tml_
and Fridrich are on Windows.... I don't think that devs will
intentionally try to break  it, but then if you care _about_ a
particular os version, you also need to care _for_ it.
iow, yes, having #ifdef to allow for different minimum OS X version
will cause some pain and some breakage, most likely on the least
used/supported version... but supporting 10.4 is causing some pain and
breakage to the rest, and particularly to  new volunteer. and it is
going to get worse as 10.4 SDK will soon vanish altogether as Joe
mentioned.

>The only effect is that you drop support for 10.4 - but I fail to see
>any positive effect (apart from the builder not having to install an
>"old" version of XCode that comes with the 10.4 SDK - but that again
>doesn't count as a reason for me)

The other effect is requiring gcc-4.0, which in turn bring its load of
pain and inefficiencies.


I suggest that we start allowing a --target-sdk= at build time for mac

Norbert