[macports-ports] branch master updated: ffmpeg: force gcc6 on PPC
Jeremy Huddleston Sequoia
jeremyhu at apple.com
Sat Oct 21 06:12:42 UTC 2017
> On Oct 20, 2017, at 20:32, Ken Cunningham <ken.cunningham.webuse at gmail.com> wrote:
>> On Oct 20, 2017, at 7:39 PM, Ryan Schmidt <ryandesign at macports.org> wrote:
>> I guess my question was: is there something unique about ffmpeg that makes it require gcc6 on PowerPC, as opposed to one of the alternatives MacPorts would choose, or is MacPorts choosing inappropriate alternatives for all ports on PowerPC? If the latter, the fix should be in base, not in individual ports.
> The latter. MacPorts is in almost all cases choosing inappropriate alternatives for PowerPC.
> On PPC, the compiler should default to gcc42 (whichever flavour - there are a couple of different variants of that - apple-gcc42 or gcc42 are for all intents and purposes the same compiler for our purposes). If *gcc42 is blacklisted, MacPorts should then go to gcc6.
The issue in doing that is that C++ ports will possibly be build inconsistently because macports-gcc6 will use the MacPorts-provided runtimes from libgcc which are not compatible with the host's libstdc++ (used by gcc42).
If you want to use gcc6 for a full build of MacPorts on ppc, I'd suggest doing something similar to the bootstrapping I documented for modern clang/libc++ on Leopard and Snow Leopard (and please document it for future users).
> There should be no other choices unless something is explicitly whitelisted (like gcc5 for pdftk).
> If you know how to make this happen, that would be great!
> [aside — that actually won’t fix quite a few ports, which just blacklist *gcc* — that has to be undone on PPC all the time, as it is inappropriate for PPC to do that.]
More information about the macports-dev