[freenet-dev] Re: [kaffe] Kaffe and Freenet, round (N+1)

[Greg Wooledge]

Toad (toad at amphibian.dyndns.org) wrote:
> On Tue, Jun 03, 2003 at 01:58:58AM +0200, Mark Wielaard wrote:
> > You might want to try to configure kaffe with --enable-pure-java-math
> > that will give you another (the GNU Classpath) java.math.BigInteger
> > implementation (which doesn't use the native libgmp).
> 
> Eeek. Sloooooooow. But I suppose if it's all that works.

OK, here we go.

I started by updating from OpenBSD 3.2 to 3.3.  Then I ran Freenet
with Kaffe from early May, which is what I've been using normally
with rather good results.  This seemed to be OK, just as it was
before the OS upgrade.

Then I tried building recent CVS Kaffe (I think it's about 2 days
old now).  I got the same results as before: "Negative bit address",
whatever that means, whenever I tried to access a document.

Then I upgraded gmp from 3.1.1 (the OpenBSD 3.2 port) to 4.1.1 (the
3.3 port).  I rebuilt Kaffe with that version of gmp, and still got
the same results: "Negative bit address".

Next, I tried with --enable-pure-java-math.  This seemed to be working
better; I could run the node and retrieve documents via fproxy (the
web interface) without "Negative bit address" errors.  But, as Toad
said, it's incredibly slow compared to gmp.  Running top, I saw
kaffe-bin take about 30-40% (with spikes up to 90%) of the CPU when
I hit a few image-heavy pages.  Normally I don't see it get over
50% even when it's doing FEC decoding -- under normal conditions,
like retrieving a few pages of images, it never got over 20% or so.

Then, after about 1:30 of CPU time, the node went into a "comatose"
state.  I see this fairly often even with the early-May Kaffe; the
node just sits there, not eating any CPU; sometimes it pulls out by
itself after a few *minutes*(!), but usually it's time to restart it.
After waiting about 5 minutes, I restarted it.

Meanwhile, stdout/stderr had this:

kaffe-bin in malloc(): warning: recursive call.
kaffe-bin in malloc(): warning: recursive call.
kaffe-bin in malloc(): warning: recursive call.

And freenet.log had this:

03-Jun-03 10:42:16 PM (freenet.interfaces.PublicInterface, Interface # tcp/36963, NORMAL): Getting thread to dispatch in PublicInterface took more than 10 seconds! If this happens frequently, report it to devl at freenetproject.org. Waited 101231 millis.
java.lang.Exception: debug
        at java.lang.Throwable.fillInStackTrace(Throwable.java:native)
        at java.lang.Throwable.<init>(Throwable.java:44)
        at java.lang.Exception.<init>(Exception.java:24)
        at freenet.interfaces.PublicInterface.dispatch(PublicInterface.java:109)
        at freenet.interfaces.Interface.acceptConnections(Interface.java:220)
        at freenet.interfaces.Interface.run(Interface.java:172)
        at java.lang.Thread.run(Thread.java:334)

(This occurred several times.)

Pretty soon, the restarted node had ground itself down into a coma
just like the first time.  top showed the CPU usage had dropped to
nothing, with about 1:33 elapsed CPU.  No more images were coming
through on the web interface.  There weren't any messages in the
logs this time, though.

So, the upshot is, I'm still running early-May CVS Kaffe.  It's the
only thing that still works.

-- 
Greg Wooledge                  |   "Truth belongs to everybody."
greg at wooledge.org              |    - The Red Hot Chili Peppers
http://wooledge.org/~greg/     |
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 187 bytes
Desc: not available
Url : http://kaffe.org/pipermail/kaffe/attachments/20030603/04d08594/attachment-0002.pgp