[KDE/Mac] Help my confusion

[Jeremy Whiting]

Wow so many choices :/

On Sun, Feb 8, 2015 at 3:08 AM, René J.V. <rjvbertin at gmail.com> wrote:

> On Saturday February 07 2015 20:11:25 Jeremy Whiting wrote:
>
> >Ok, googling showed me this:
> >
> https://trac.macports.org/browser/trunk/dports/lang/hugs98/files/patch-packages-base-include-HsBase.h.diff?rev=81676
> >which fixes kio also. Does it make sense to check that into kio itself
> with
>
> It's quite possible that that's the same fix I already had to apply at
> some time in the past. Makes sense, right - you get a complaint about a
> symbol that clearly is "standard" and should be a global ... try declaring
> it extern with the appropriate type and see how that goes :)
>
> Reminds me of a C++11 specific typedef that was missing in clang 3.5 until
> not so long ago; this kind of issue is probably due to an oversight from
> the compiler developers.
>

After reading man environ on both systems I've submitted a review request
that builds on both machines also https://git.reviewboard.kde.org/r/122481/


>
> >an #ifdef Q_OS_MAC, or is that not needed anymore on "modern" OS X? I'll
>
> It would be cleaner to put it in a block that depends on the OS X version
> (you'll need AvailabilityMacros.h for that).
>
> As to C++11:
> > another problem, apparently std doesn't have lround which kdeclarative
> uses
> > :/. Apparently it's from C++11, is there a C++11 standard library on OSX
> > Lion or is that too old of a system to have such a new set of libraries?
>
> I keep to my personal rule of thumb that C++11 support isn't complete in
> clang on pre-10.8 or 10.9 systems because it depends on system headers, but
> the situation is more subtle than that IIRC. On 10.6.8 I came to the
> conclusion that it was better to use configure.compiler=macports-gcc-4.9 at
> once when I noticed the use of C++11. That's how I could build KDevelop 4.7
> . It's not guaranteed NOT to lead to conflicts but in practice that never
> happened to me.
>

What compiler is used by macports, or is that configurable by the end user?
Another idea I have is could we ifdef for C++11 support and use Qt::round
or something else when C++11 isn't available on a give platform? Is there
something we could #ifdef on to get that? iirc KDE Applications can opt to
use C++11 if they want, but I thought it wasn't permitted in libraries (so
this use in kdeclarative maybe should get removed anyway, not sure.)

>
> On a side-note: using gcc also gave a nice and useful speed boost. I think
> I've recently provided some evidence that even on 10.9 macports-clang-3.5
> is slower (just 2x or so) than the almost identical AppleClang version; on
> 10.6 and presumably 10.7 you don't have that choice but FOSS gcc ought to
> be perfectly usable as long as you don't require Apple specific features.
> Note that by using a dev. compiler version you're exposing yourself to
> unforeseen issues but also to the fact that MacPorts builds it (and version
> 3.6. too, I think, still) with assertions enabled, which degrades
> performance even more.
>
> As to performance: clang does have a nice feature in that it's inherently
> a cross-compiler that can generate code for different targets. On OS X, it
> can also generate gnu-linux object code, and on Linux it can also generate
> mach-o object codes. Setting it up for use with distcc in a heterogenous
> cluster is thus a matter of writing a few wrappers that add the appropriate
> output option, and figuring out how MacPorts' support for distcc works.
> Sadly that won't help if you have a C++11 problem...
>

Cross compling and using distcc may be interesting in the future, first I'd
like to get everything to build on the platforms I have though.

thanks,
Jeremy

>
> R.
> _______________________________________________
> kde-mac at kde.org
> List Information: https://mail.kde.org/mailman/listinfo/kde-mac
> KDE/Mac Information: http://community.kde.org/Mac
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/kde-mac/attachments/20150208/8d35cc7f/attachment.html>