How to debug Kaffe?

Erik Schnetter erik.schnetter at
Wed May 14 00:18:03 PDT 1997


I am trying to use kaffe (version 0.84) for some code I compiled using a
different compiler that javac.  Of course I ran into some problems, and I
have come to think that kaffe might not be entirely innocent.  This leads
me to the question: How do I debug kaffe itself?  (I am using kaffe on

I am aware of the options to trace method calls, and to print the address
of every function compiled to native code.  I can change my source code to
print some progress and/or debug information.  However, the bug seems to
be somewhat persistent, and I'd like some more comfort. 

Is there a way to set breakpoints, e. g. using gdb?  Is there a
"-breakpoint" option to set a breakpoint in a specific function, or is
there a way to at least to (say) raise a signal at a specific time?  (I
could catch that signal in a debugger, or ignore that signal at will.) Is
there any debug information available in kaffe-compiled code?  Another
question: is there a way to access the floating point registers from gdb,
or from another debugger on Linux? 

The bug I'm encountering (rather: the bug that haunts me) has to do with
single precision floating point arithmetic (double precision works ok) and
occurs only if I execute some other (though apparently totally unrelated)
statements in a certain order before.  The bytecode compares a floating
point value to itself with the result being "not equal".  This register
should contain the value zero (and not NaN) -- although you never know. 

So what I'm really still trying to do is to find out what this bug is
about.  In a similar situation some time ago (but not with kaffe) the
reason turned out to be inconsistent usage of the floating point unit
(across several indedepently compiled C files) where the floating point
unit stack got "confused".  (A function returning a single-precision value
had no prototype, so the returned float was interpreted as
double-precision value.  C is, after all, not typesafe.  But that's
another story and shall be told another time.)

Or did someone have similar experiences?


Erik Schnetter, erik.schnetter at

Arthur C. Clarke saw the future.
	He knew it would be run by huge computers.
		And he was right.		Just not about the size.

More information about the kaffe mailing list