[plt-scheme] preliminary "Snow Leopard" report
The version of the cgc (bohem) collector included in mzscheme is
somewhat dated, it may finally be time to sync up with a more current
version. Parallel collection on x86_64 doesn't work either.
Kevin
John Clements wrote:
>
> On Sep 4, 2009, at 7:27 PM, John Clements wrote:
>
>>
>> On Sep 4, 2009, at 6:02 PM, John Clements wrote:
>>
>>> Trying to run mzscheme after installing 10.6 gives this error:
>>>
>>> You need the Rosetta software to run mzscheme. The Rosetta installer
>>> is in Optional Installs on your Mac OS X installation disc.
>>> Schmoop-Box-3:~ clements$
>>>
>>>
>>> I haven't even tried googling for this error to see what the problem
>>> is; it might be as simple as recompiling.
>>
>> Yep; leftover PPC code. Looks like a "please ignore this message."
>
> Okay, more fun. A clean build (with the new version of XCode, 3.2) of
> the svn head gives first a simple error:
>
> cd gc; make ../libmzgc.a
> gcc -I../../../mzscheme/gc/include -DNO_SIGNALS -g -O2 -Wall -DOS_X
> -D_DARWIN_UNLIMITED_SELECT -fno-common -DNO_EXECUTE_PERMISSION
> -DSILENT -DNO_GETENV -DLARGE_CONFIG -DATOMIC_UNCOLLECTABLE
> -DINITIAL_MARK_STACK_SIZE=8192 -c ../../../mzscheme/gc/alloc.c
> In file included from ../../../mzscheme/gc/include/private/gc_priv.h:66,
> from ../../../mzscheme/gc/alloc.c:19:
> ../../../mzscheme/gc/include/private/gcconfig.h:476: error: expected
> identifier or ‘(’ before ‘--’ token
> ../../../mzscheme/gc/alloc.c: In function ‘GC_expand_hp_inner’:
> ../../../mzscheme/gc/alloc.c:966: warning: suggest parentheses around
> && within ||
> make[6]: *** [alloc.o] Error 1
> make[5]: *** [g-c] Error 2
> make[4]: *** [common] Error 2
> make[3]: *** [cgc] Error 2
> make[2]: *** [3m] Error 2
> make[1]: *** [3m] Error 2
> make: *** [all] Error 2
>
> ... This turns out to be because the code in gcconfig.h doesn't have a
> case for this system, apparently. Since I don't have a pre-10.6
> machine to experiment with, I guessed that we wanted to wind up in the
> 32-bit intel macosx bin, and I changed
>
> # if defined(__i386__)
> # define I386
>
> into
>
> # if defined(__i386__) || defined(__x86_64__)
> # define I386
>
> Unfortunately, the mzschemecgc that results from this dumps core:
>
> env XFORM_PRECOMP=yes ../mzschemecgc -cqu
> ../../../mzscheme/gc2/xform.ss --setup . --cpp "gcc -E -I./..
> -I../../../mzscheme/gc2/../include -DOS_X -D_DARWIN_UNLIMITED_SELECT
> -fno-common -DNEWGC_BTC_ACCOUNT " --keep-lines -o xsrc/precomp.h
> ../../../mzscheme/gc2/precomp.c
> make[4]: *** [xsrc/precomp.h] Segmentation fault
> make[3]: *** [all] Error 2
> make[2]: *** [3m] Error 2
> make[1]: *** [3m] Error 2
> make: *** [all] Error 2
>
>
> So, just for fun, I tried adding a new #ifdef clause to special-case
> the x86_64:
>
> # if defined(__x86_64__)
> # define X86_64
> # define mach_type_known
> # endif
>
> ... but this leads to an error about an undefined variable DATASTART:
>
> ^^^^Starting command^^^^
> gcc -I../../../mzscheme/gc/include -DNO_SIGNALS -g -O2 -Wall -DOS_X
> -D_DARWIN_UNLIMITED_SELECT -fno-common -DNO_EXECUTE_PERMISSION
> -DSILENT -DNO_GETENV -DLARGE_CONFIG -DATOMIC_UNCOLLECTABLE
> -DINITIAL_MARK_STACK_SIZE=8192 -c ../../../mzscheme/gc/os_dep.c
> ../../../mzscheme/gc/os_dep.c: In function ‘GC_get_stack_base’:
> ../../../mzscheme/gc/os_dep.c:1069: error: ‘result’ undeclared (first
> use in this function)
> ../../../mzscheme/gc/os_dep.c:1069: error: (Each undeclared identifier
> is reported only once
> ../../../mzscheme/gc/os_dep.c:1069: error: for each function it
> appears in.)
> ../../../mzscheme/gc/os_dep.c: In function ‘GC_register_data_segments’:
> ../../../mzscheme/gc/os_dep.c:1457: error: ‘DATASTART’ undeclared
> (first use in this function)
> make[6]: *** [os_dep.o] Error 1
> make[5]: *** [g-c] Error 2
> make[4]: *** [common] Error 2
> make[3]: *** [cgc] Error 2
> make[2]: *** [3m] Error 2
> make[1]: *** [3m] Error 2
> make: *** [all] Error 2
>
> ... which suggests to me (per earlier plt-scheme discussions) that
> mzscheme + OS X + 64-bit is not working.
>
> FWIW, the downloaded intel binary from plt-scheme.org still works just
> fine.
>
> Again, I don't have a 10.5 machine to see exactly what bin it's
> *supposed* to wind up in, so there may be a simple fix for this.
>
> All the best,
>
> John
>
>
>
> _________________________________________________
> For list-related administrative tasks:
> http://list.cs.brown.edu/mailman/listinfo/plt-scheme