Re: [fedora-java] Re: java issues on ppc
Orion Poplawski writes:
Robin Green wrote:
> It's tricky - but possible - to debug interpreted bytecode. JDWP protocol
> support is being worked on, but it's not ready yet.
>
> Is plplot multithreaded? I've just discovered a thread-safety bug in the
> gij interpreter which has lain unfixed for several years:
>
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23367
>
> I'm told that this bug only affects the interpreter.
>
> There are also a couple of other threading bugs that I know of (one of
> which is not so easy to reproduce on x86 but easy to reproduce on x86-64,
> FWIW). But both of those are not specific to interpreted code.
>
Well, plplot is compiled with threading, but I'm not sure it make much
use of them. I'll try compiling without threading to see if that makes
a difference.
> If you want to join the #fedora-java IRC channel on
> irc.freenode.net, I can try and help you debug it over IRC.
>
I'll try soon...
> What kind of problems are you experiencing on ppc, when running plplot as
> interpreted?
>
Basically, I've got the following call path:
main:
pls.w3d( 1.0, 1.0, 1.0, -1.5, 1.5, -0.5, 1.5, zmin, zmax, alt[k], az[k] );
->
package plplot.core;
class plplotjavacJNI {
public final static native void plw3d(double jarg1, double jarg2,
double jarg3, double jarg4, double jarg5, double jarg6, double jarg7,
double jarg8, double jarg9, double jarg10, double jarg11);
}
If I print out the values of the arguments in plplotjavac.plw3d, some of
them are bogus (i.e. basex = 0.0 when it is passed 1.0). Others are okay.
Ah, right. This m,ight well be a libffi problems that is specific to
PPC.
Andrew.