|
|
| Ron Garret 2007-05-09, 4:11 am |
| Has anyone successfully built the Boehm GC on OS X 10.4? I've tried
every version up to and including 7.0alpha7 and they all fail the self
test. (I'm running on a Macbook Pro, i.e. a Core2 duo.)
Failing that, are there any other GC libraries floating around out there
that might work instead?
Thanks,
rg
| |
| llothar 2007-05-09, 4:11 am |
| On 9 Mai, 05:03, Ron Garret <rNOSPA...@flownet.com> wrote:
> Has anyone successfully built the Boehm GC on OS X 10.4? I've tried
> every version up to and including 7.0alpha7 and they all fail the self
> test. (I'm running on a Macbook Pro, i.e. a Core2 duo.)
>
> Failing that, are there any other GC libraries floating around out there
> that might work instead?
You need to get the CVS version, there were a lot of hacks for MacOSX
in the past.
Unfortunately it depends also on the XCode version you use. There are
a few problems (caused by the Apple side), in that there isn't a 100%
reliable way to get the CPU registeres. They changed the API for this
a few times.
I haven't seen any other reliable plug-in GC. You need to google a
litte bit, because a few years back there was an announcement about a
reliable mark-and-copy collector but it needs a lot of help from the
code generator.
| |
| Juanjo 2007-05-09, 10:04 pm |
| On May 8, 3:03 pm, Ron Garret <rNOSPA...@flownet.com> wrote:
> Has anyone successfully built the Boehm GC on OS X 10.4? I've tried
> every version up to and including 7.0alpha7 and they all fail the self
> test. (I'm running on a Macbook Pro, i.e. a Core2 duo.)
ECL uses the version 6.8 of the garbage collector.
Juanjo
| |
| Rayiner Hashem 2007-05-10, 4:17 am |
| > You need to get the CVS version, there were a lot of hacks for MacOSX
> in the past.
Are you sure? The latest 6.x series release works just fine on my
MacBook. I've been using it in OpenDylan for several months with no
incident.
| |
| Ron Garret 2007-05-10, 10:04 pm |
| In article <1178775575.502187.21240@q75g2000hsh.googlegroups.com>,
Rayiner Hashem <rayiner@gmail.com> wrote:
>
> Are you sure? The latest 6.x series release works just fine on my
> MacBook. I've been using it in OpenDylan for several months with no
> incident.
What flavor of Macbook are you using? Does it pass the self-tests?
rg
| |
| Rayiner Hashem 2007-05-10, 10:04 pm |
| > What flavor of Macbook are you using? Does it pass the self-tests?
>
> rg
I just noticed that you have to build with --disable-threads to allow
"make check" to pass. I guess I never noticed it since I don't use it
multithreaded.
| |
| Ron Garret 2007-05-10, 10:04 pm |
| In article <1178845125.982448.281520@y5g2000hsa.googlegroups.com>,
Rayiner Hashem <rayiner@gmail.com> wrote:
>
> I just noticed that you have to build with --disable-threads to allow
> "make check" to pass. I guess I never noticed it since I don't use it
> multithreaded.
Thanks!!!
rg
| |
| Ron Garret 2007-05-11, 10:04 pm |
| In article <1178845125.982448.281520@y5g2000hsa.googlegroups.com>,
Rayiner Hashem <rayiner@gmail.com> wrote:
>
> I just noticed that you have to build with --disable-threads to allow
> "make check" to pass. I guess I never noticed it since I don't use it
> multithreaded.
Well, that worked, but it uncovered another problem:
gc/new_gc_alloc.h includes bits/stl_alloc.h which no longer exists in
gcc4, and gcc3 is no longer supported on Intel Macs.
I tried a few naive workarounds, but none of them worked.
:-(
rg
| |
| Ron Garret 2007-05-11, 10:04 pm |
| In article <rNOSPAMon-F69389.10495911052007@news.gha.chartermi.net>,
Ron Garret <rNOSPAMon@flownet.com> wrote:
> In article <1178845125.982448.281520@y5g2000hsa.googlegroups.com>,
> Rayiner Hashem <rayiner@gmail.com> wrote:
>
>
> Well, that worked, but it uncovered another problem:
>
> gc/new_gc_alloc.h includes bits/stl_alloc.h which no longer exists in
> gcc4, and gcc3 is no longer supported on Intel Macs.
>
> I tried a few naive workarounds, but none of them worked.
Update: it appears that bits/stl_alloc.h is being included in order to
get the simple_alloc template, which seems to have completely vanished
from gcc4:
[ron@mickey:/usr/include/c++/4.0.0]$ grep -r simple_ .
[ron@mickey:/usr/include/c++/4.0.0]$
rg
| |
| Rayiner Hashem 2007-05-11, 10:04 pm |
| As far as I can tell, new_gc_alloc.h is a special header for using SGI
STL's non-standard extensions. Most such extensions were removed in
GCC 4.0. Is there any particular reason you need to use this header?
Also, GCC 3.x (and all its headers) is still in Tiger, it's just named
g++-3.3.
> Update: it appears that bits/stl_alloc.h is being included in order to
> get the simple_alloc template, which seems to have completely vanished
> from gcc4:
>
> [ron@mickey:/usr/include/c++/4.0.0]$ grep -r simple_ .
> [ron@mickey:/usr/include/c++/4.0.0]$
>
> rg
| |
| Ron Garret 2007-05-11, 10:04 pm |
| In article <1178912928.391309.25380@o5g2000hsb.googlegroups.com>,
Rayiner Hashem <rayiner@gmail.com> wrote:
> As far as I can tell, new_gc_alloc.h is a special header for using SGI
> STL's non-standard extensions. Most such extensions were removed in
> GCC 4.0. Is there any particular reason you need to use this header?
Well, yes and no. I'm trying to compile some code I wrote a long time
ago on Linux which uses the SGI STL API. (How's that for some TLA's?)
I suppose I could try to rewrite it using the more standard interface,
but I was really hoping to avoid that.
> Also, GCC 3.x (and all its headers) is still in Tiger, it's just named
> g++-3.3.
Apparently it doesn't work:
[ron@mickey:~/devel/ciel]$ g++-3.3 reader.cc
g++-3.3: installation problem, cannot exec `cc1plus': No such file or
directory
This is after a fresh install of XCodeTools2.4.1 (and even after a
separate re-install of the GCC3.3 packaage). I've found some stuff on
the Web that indicates that 3.3 is not supported on Intel macs.
rg
| |
| Tim Bradshaw 2007-05-11, 10:04 pm |
| On 2007-05-11 21:09:11 +0100, Ron Garret <rNOSPAMon@flownet.com> said:
> How's that for some TLA
SGI INA TLA IMO, though it probably was when you wrote the code :-)
| |
| Rayiner Hashem 2007-05-11, 10:04 pm |
| Tricky problem. You could always try STLport (5.x), which seems to
implement this particular extension, and is apparently supported on
Darwin.
| |
|
|
| Ron Garret 2007-05-12, 10:07 pm |
| In article <1178926606.786640.310100@l77g2000hsb.googlegroups.com>,
Rayiner Hashem <rayiner@gmail.com> wrote:
> Tricky problem. You could always try STLport (5.x), which seems to
> implement this particular extension, and is apparently supported on
> Darwin.
I think I'll just update my code to use the new interface. The SGI
stuff is getting pretty stale.
rg
| |
|
|
|
|