[kaffe] total NetBSD/ppc breakage

[Dalibor Topic]

Riccardo wrote:
> Hello,
> 
> When compiling kaffe (today's CVS, clean check out) on NetBSD/ppc (2.0 
> beta, gcc 3.3.3) I get:
> gmake[2]: Leaving directory `/home/multix/kaffe-cvs/netbsd-ppc/kaffe/
> kaffeh'
> ./kaffe/kaffeh/kaffeh -classpath ../../kaffe/libraries/javalib/Klasses.
> jar.bootstrap -o java_lang_Object.h java/lang/Object
> assertion "cf != NULL" failed: file "../../../kaffe/kaffe/kaffevm/file.
> h", line 116, function "readu2"
> [1]   Abort trap (core dumped) ../kaffe/kaffeh/...
> gmake[1]: *** [stamp-h0all] Error 134
> 
> 
> when trying to compile without inline functions.... the error doesn't 
> exhibit and I can go much much further, up to:
> cc1: warning: "-fno-default-inline" is valid for C++ but not for C/ObjC
> cc1: warning: -Wuninitialized is not supported without -O

Okay, then I'll do plan B here, and move the functions to their own file.

Class file handling should be in its own shared library eventually, 
since it's used both by the vm and by kaffeh. The current build system 
with partially overwriting files in kaffeh is rather weird. Anyone knows 
a good IDE with refactoring support for C? :)

> In file included from ../../../../../kaffe/config/powerpc/netbsd1/md.h:
> 15,
>                  from ../../../../config/md.h:1,
>                  from ../../../../../kaffe/kaffe/kaffevm/systems/unix-
> jthreads/jthread.h:41,
>                  from ../../../../../kaffe/kaffe/kaffevm/systems/unix-
> jthreads/signal.c:18:
> ./../../../../kaffe/config/powerpc/common.h:21:5: warning: "NEED_
> sysdepCallMethod" is not defined
> ./../../../../kaffe/kaffe/kaffevm/systems/unix-jthreads/signal.c:38: 
> error: syntax error before '(' token
> ./../../../../kaffe/kaffe/kaffevm/systems/unix-jthreads/signal.c: In 
> function `jthread_initexceptions':
> ./../../../../kaffe/kaffe/kaffevm/systems/unix-jthreads/signal.c:60: 
> error: `nullException' undeclared (first use in this function)
> ./../../../../kaffe/kaffe/kaffevm/systems/unix-jthreads/signal.c:60: 
> error: (Each undeclared identifier is reported only once
> 
> 
> so, this is strange, maybe a GCC bug? or something wicked in the code 
> that confuses gcc.

No, I think Guilhem just forgot for add SIGNAL_ARGS to powerpc ports 
other than darwin :) I've CC:ed him on this mail, hopefully he has some 
time to look into it.

cheers,
dalibor topic