Adam Thomason <[EMAIL PROTECTED]> wrote: > Inspired by Andy Dougherty's Sun exercise, I've taken a crack at > building parrot on an IBM RS/6000 running AIX 4.3 with VisualAge C 6. > Happily, it almost works. Issues I found:
Great, thanks. > * xlc had the same problem with bit.ops as Sun's compiler. Luke Palmer's change > fixes the compile error, and the matching segfault in bitwise test 7. > * Andy's patch to pmc2c.pl was also necessary. These 2 got already committed. > * Vtable.pm had an extra comma in an initializer block on line 76. Yep. > * xlc seemingly has no facility for inlining assembly code, so I just > commented out the __asm__ bits in ppc_sync_cache (jit/ppc/jit_emit.h) > to let it compile; this is quite obviously wrong. Seems, that you have to roll your own assembler source file and we need some config support to include that in the build. This could probably be done in feature_h.pl, but a generic scheme to assemble and link asm would be fine. > * -lpthreads is required in libs. I just put it in my config/init/hints/local.pl. That's missing in the patch. You should create "$^O.pl" in hints though. > * Some tests in io.t needed to include parrot/embed.h. > I've rolled up the important changes above into a patch against current CVS. > The one remaining test failure in 'make test' is op/number test 10, > the second part of which computes mod(0.0,3.0). The expected value is > 0.0, but I see -0.0. However, given the definition of > mod(x,y)=x-y*floor(x/y), this case reduces to 0.0-0.0*0.0, which could > justify -0.0. Perhaps the test can be relaxed somehow to accomodate > minus-zero semantics. Sounds correct, yes. > 'make testj' still fails 863/942 subtests, including even basic_1 > (noop). I'm not sure how to debug this, as the debugger won't step > inside the call to core(interpreter, pc) in interpreter.c. Does your debugger support stabs? You could try to enable the jit_debug stuff in jit.c and follow the hints in docs/jit.pod /Debugging. > Adam Thomason Thanks, applied missing pieces. leo
