Thread overview
phobos compile error on PPC G5
Jan 07, 2005
David Friedman
Jan 07, 2005
David Friedman
January 06, 2005
Just testing a PPC64 compile of
GDC 0.9 using GCC 3.4.3 sources:

> gdc -o std/dateparse.o -Os -pipe -mcpu=G5 -frelease  -nostdinc -I ../gcc/d/phobos -I ../gcc/d/phobos/internal/gc -c ../gcc/d/phobos/std/dateparse.d
> d/dmd/todt.c:148: failed assertion `sz <= v->type->size()'
> ../gcc/d/phobos/std/dateparse.d: In member function `classify':
> ../gcc/d/phobos/std/dateparse.d:277: internal compiler error: Abort trap
> Please submit a full bug report,
> with preprocessed source if appropriate.
> See <URL:http://gcc.gnu.org/bugs.html> for instructions.
> {standard input}:155:non-relocatable subtraction expression, "L__init_3std4date4Date$non_lazy_ptr" minus "L00000000002$pb"
> {standard input}:155:symbol: "L__init_3std4date4Date$non_lazy_ptr" can't be undefined in a subtraction expression
> {standard input}:149:non-relocatable subtraction expression, "L__init_3std4date4Date$non_lazy_ptr" minus "L00000000002$pb"
> {standard input}:149:symbol: "L__init_3std4date4Date$non_lazy_ptr" can't be undefined in a subtraction expression
> {standard input}:unknown:Undefined local symbol L__d_arraycatn$stub
> {standard input}:unknown:Undefined local symbol L__D6object5Error5_ctorFAaZC6object5Error$stub
> {standard input}:unknown:Undefined local symbol L__init_3std4date4Date$non_lazy_ptr
> {standard input}:unknown:Undefined local symbol L__d_newclass$stub
> {standard input}:unknown:Undefined local symbol L__d_throw$stub
> gnumake[1]: *** [std/dateparse.o] Error 1

Mac OS X 10.3, PowerPC G5 (970FX)
DFLAGS="-Os -pipe -mcpu=G5 -frelease"

--anders
January 07, 2005
Anders F Björklund wrote:
> Just testing a PPC64 compile of
> GDC 0.9 using GCC 3.4.3 sources:
> 
>> gdc -o std/dateparse.o -Os -pipe -mcpu=G5 -frelease  -nostdinc -I ../gcc/d/phobos -I ../gcc/d/phobos/internal/gc -c ../gcc/d/phobos/std/dateparse.d
>> d/dmd/todt.c:148: failed assertion `sz <= v->type->size()'
>> ../gcc/d/phobos/std/dateparse.d: In member function `classify':
>> ../gcc/d/phobos/std/dateparse.d:277: internal compiler error: Abort trap
>> Please submit a full bug report,
>> with preprocessed source if appropriate.
>> See <URL:http://gcc.gnu.org/bugs.html> for instructions.
>> {standard input}:155:non-relocatable subtraction expression, "L__init_3std4date4Date$non_lazy_ptr" minus "L00000000002$pb"
>> {standard input}:155:symbol: "L__init_3std4date4Date$non_lazy_ptr" can't be undefined in a subtraction expression
>> {standard input}:149:non-relocatable subtraction expression, "L__init_3std4date4Date$non_lazy_ptr" minus "L00000000002$pb"
>> {standard input}:149:symbol: "L__init_3std4date4Date$non_lazy_ptr" can't be undefined in a subtraction expression
>> {standard input}:unknown:Undefined local symbol L__d_arraycatn$stub
>> {standard input}:unknown:Undefined local symbol L__D6object5Error5_ctorFAaZC6object5Error$stub
>> {standard input}:unknown:Undefined local symbol L__init_3std4date4Date$non_lazy_ptr
>> {standard input}:unknown:Undefined local symbol L__d_newclass$stub
>> {standard input}:unknown:Undefined local symbol L__d_throw$stub
>> gnumake[1]: *** [std/dateparse.o] Error 1
> 
> 
> Mac OS X 10.3, PowerPC G5 (970FX)
> DFLAGS="-Os -pipe -mcpu=G5 -frelease"
> 
> --anders

The -mcpu=G5 option (-mpowerpc64, actually) is triggering the problem. This creates an interesting situation in which the word size is 64 bits, but it's really still 32 bits :-)

David
January 07, 2005
David Friedman wrote:

> The -mcpu=G5 option (-mpowerpc64, actually) is triggering the problem. This creates an interesting situation in which the word size is 64 bits, but it's really still 32 bits :-)

Can you not compile a PPC64 application, that can run on Mac OS X 10.3 ?
I thought you could, http://developer.apple.com/technotes/tn/tn2086.html

I know that the entire program userland and adress space is 32-bit,
I just wanted to play with native longlongs and the other stuff...

Guess I will have to wait for: a) Tiger or b) decent Gentoo fans, then

--anders
January 07, 2005
Anders F Björklund wrote:
> David Friedman wrote:
> 
>> The -mcpu=G5 option (-mpowerpc64, actually) is triggering the problem. This creates an interesting situation in which the word size is 64 bits, but it's really still 32 bits :-)
> 
> 
> Can you not compile a PPC64 application, that can run on Mac OS X 10.3 ?
> I thought you could, http://developer.apple.com/technotes/tn/tn2086.html
> 
> I know that the entire program userland and adress space is 32-bit,
> I just wanted to play with native longlongs and the other stuff...
> 
> Guess I will have to wait for: a) Tiger or b) decent Gentoo fans, then
> 
> --anders

It's just a small bug I need to fix.  To build Phobos with -mpowerpc64, however, I would also need to add the capibility to distinguish between the "machine word size" and the "ABI word size".  Actually, only two modules care about this: the exception handling routines and the Mach interface.

I think you would also have problems if you build libstdc++ or libgcc with -mpowerpc64.

David