[kaffe] 1.1.0 on alpha-linux: make check crashes

[Dalibor Topic]

Hi Kevin,

--- "Kevin D. Kissell" <kevink at mips.com> wrote:
> > > Could you please check confirm that the 4 regression tests
> > > which succeed on alpha-linux are the same 4 that succeed
> > > with MIPS/Linux under jit3, i.e.
> > >
> > > HelloWorldApp.class.save
> > > CatchLimits.class.save
> > > TestFloatDouble, and
> > > ThreadStop?
> > 
> > FYI, these tests almost always succeed (They have no output to check 
> > against so they pass by default).  I think the only way they can fail 
> > is if the executable exits with an error code.
> 
> But what's intriguing here is that TestFloatDouble and
> ThreadStop are being successfully *compiled*, wheras
> all other test/regression Java files are causing JIT3 to
> explode on RISC platforms during their compilation.
> I agree that, once they've been compiled, it's hardly
> surprising that they do not fail in execution!
> 

I took another look at this one and it may shed some light on the problems you
were seing:

/tmp/kaffe/install/bin/kaffe -vmdebug ALL at.dms.kjc.Main HelloWorldApp.java
2>out
[...]
tidyAnalyzeMethod 0x1200da030: clearing codeInfo 0x11fffeda8
Translated
java/lang/ClassLoader.findLoadedClass(Ljava/lang/String;)Ljava/lang/Class;
(normal) 0x120246be8
enable stop for thread  0x1200da030
enable stop for thread  0x1200da030
soft_fixup_trampoline(): return 0x120246be8
on current stack: base=0x11ff00000 size=1048576 bp=0x11fffcd70 yes
on current stack: base=0x11ff00000 size=1048576 bp=0x11fffcc40 yes
on current stack: base=0x11ff00000 size=1048576 bp=0x11fffcb10 yes
on current stack: base=0x11ff00000 size=1048576 bp=0x11fffc9d0 yes
on current stack: base=0x11ff00000 size=1048576 bp=0x11fffc8a0 yes
on current stack: base=0x11ff00000 size=1048576 bp=0x11fffc760 yes
on current stack: base=0x11ff00000 size=1048576 bp=0x11fffc890 yes
soft_fixup_trampoline(): meth 0x120252918, where 0x120252930
disable stop for thread  0x1200da030
  0x1200da030 entering process class kaffe/lang/PrimordialClassLoader 7->12
findMethodLocal(kaffe/lang/PrimordialClassLoader,finalize,()V) -> NOT FOUND
findMethodLocal(java/lang/ClassLoader,finalize,()V) -> NOT FOUND
findMethodLocal(java/lang/Object,finalize,()V) -> 0x12011c8b8
enable stop for thread  0x1200da030
disable stop for thread  0x1200da030
Initialising kaffe/lang/PrimordialClassLoader static 88
findMethodLocal(kaffe/lang/PrimordialClassLoader,<clinit>,()V) -> 0x120252c78
enable stop for thread  0x1200da030
using JNI
firstFrame pc 0x20000050738 sp 0x11fffebb0 fp (nil)
on current stack: base=0x11ff00000 size=1048576 bp=0x11fffebb0 yes
lda $sp,112($sp) at 0x20000050d78
ret $31,($26),1 at pc 0x20000050d7c
ldq $fp,56($sp) at 0x20000050d74
ldq $26,0($sp) at 0x20000050d58
kaffe-bin: ../../../kaffe-1.1.0/kaffe/kaffevm/exception.h:150:
vmExcept_setJNIFrame: Assertion `fp != 0' failed.

there is one interesting warning from the compiler for the jitter:
../../../../kaffe-1.1.0/kaffe/kaffevm/jit/icode.c: In function
`softcall_lookupinterfacemethod':
../../../../kaffe-1.1.0/kaffe/kaffevm/jit/icode.c:3307: warning: cast to
pointer from integer of different size

it seems to crash after jitting the clinit method of the PrimordialClassLoader.
gdb claims stack corruption ;)

cheers,
dalibor topic

__________________________________
Do you Yahoo!?
Yahoo! Calendar - Free online calendar with sync to Outlook(TM).
http://calendar.yahoo.com