I'm using gcc 4.0.1 on 10.5.x.
It should be easy enough to undefine it for pthreads.cc but under what
condition? For a linux gnu build it should still be defined correct?
How much of a speedup is sincos anyway can we just take that out?
Carson
On Mon, 2009-04-06 at 10:58 -0600, Thiago Ize wrote:
_GNU_SOURCE should not be defined on macs, which is why I did not
include the __APPLE__ guard. However, after greping the code, I see a
Core/Thread/Thread_pthreads.cc:# define _GNU_SOURCE
I guess if that file gets compiled before the the Trig.h file we get
the problem. When I compile manta on my mac, when Trig.h gets compiled
_GNU_SOURCE is not yet defined which is why it worked for me.
So either we need to undefine _GNU_SOURCE after using it in
Thread_pthreads.cc or we need to place extra guards in Trig.h. Extra
guards in Trig.h though is a bit of a hack and won't work on all
systems (non-apple systems that shouldn't have _GNU_SOURCE defined or
future apple systems that can use it)
Anyone have any preferences for which fix to use?
Thiago
Solomon Boulos wrote:
I updated and built recently (whenever I sent out mail about Apple's
broken update to libfreetype).
What compiler are you using? If gcc, what does gcc --version spit
out? If icc, they play weird games with pretending to be GCC-like,
and perhaps this is a pitfall?
Solomon
On Apr 6, 2009, at 8:37 AM, James Bigler
< >
wrote:
Has this recently changed. I've compiled Manta a few months ago
on my Mac (10.5). Could this be a 10.4 issue?
James
On Mon, Apr 6, 2009 at 2:11 AM, Carson Brownlee
< >
wrote:
I've noticed that sincos in Core/Math/Trig.h breaks under
my Mac laptop. As I understand it _GNU_SOURCE should make
math.h include sincos which seems to work fine on various
linux machines I have used. Is this a problem across all
macs and a simple !(__APPLE__) will fix this or is this
specific to my laptop?
Carson
Archive powered by MHonArc 2.6.16.