svn commit: r395079 - in head/graphics: . mitsuba mitsuba/files

[Alexey Dokuchaev]

On Mon, Aug 24, 2015 at 11:01:04AM +0200, Baptiste Daroussin wrote:
> On Mon, Aug 24, 2015 at 10:48:07AM +0200, Baptiste Daroussin wrote:
> > On Mon, Aug 24, 2015 at 07:09:15AM +0000, Alexey Dokuchaev wrote:
> > > [...]
> > > That said, OSVERSION < 900014 will not always be false.  And why do you
> > > want to be rude to our users and deliberately break things just because
> > > we no longer promise that they will work?
> > 
> > Even in that case, this is wrong anyway as you enforce GCC while should
> > should require a modern compiler via compiler.mk.

I've thought of Uses/compiler.mk, but eventually couldn't decide which
compiler to ask for: the problem in my case was not with the code (so it
is not like "port needs a compiler understanding C++0X/C++11/C++14", but
the bugginess of GCC 4.2.1 itself.  (Now that I'm reading through it again,
compiler:openmp looks interesting).

Or there's a way to require a modern compiler thish is not feature-based?

> Btw that would even be nicer to only use compiler.mk feature as it will
> give this port a chance to work on supported freebsd version on some
> tiers-2 architecture which depends on gcc 4.2.1: mips, powerpc, sparc64,
> etc. While your patch only gives a chance for it to work on unsupported
> version :)

Certainly, and usually I try to be careful about tiers-2.  In this case,
however, I'd have to find a way to de-SSE2'ize the code beforehand; it is
on my list, but quite of low priority.

./danfe