Re: [kaffe] (sigh) Kaffe 1.1 can't do FEC decoding

2003-06-13 Thread Greg Wooledge
Dalibor Topic ([EMAIL PROTECTED]) wrote:

> Hm, any chance of pinpointing the exact checkin on kaffe that started failing
> FEC ? kaffe -fullversion of the last version that works, kaffe -fullversion of
> the first version that fails, and an ugly binary search over checkouts, I guess
> ;) There have been quite a few changes to the core libraries in the last 4
> weeks.

In each case below, I checked out CVS, removed the /usr/local/kaffe
symlink, did a "./configure --enable-debug --with-includes=/usr/local/include
--with-libraries=/usr/local/lib", then "vi libtool", "gmake", "gmake install",
then renamed /usr/local/kaffe to /usr/local/kaffe-May-NN and made
/usr/local/kaffe a symlink to kaffe-May-NN.  Then I renamed "servlet.jar"
within /usr/local/kaffe-May-NN/jre/lib to "disabled.servlet.jar"
and did my testing.  After testing, I did a "gmake distclean" before
the next checkout.

From the freenet point of view, what I did was create a new user
account, drop "freenet.jar" and "freenet-ext.jar" and "seednodes.ref"
into its home dir, ran "java freenet.node.Main --config" and set it
up as a transient on a randomly chosen port with defaults for
everything else.  I inserted a 2.3 MB file using
"java freenet.client.cli.Main --htl 0 put CHK@ filename", and then
with each version of Kaffe I attempted to retrieve it with
"java freenet.client.cli.Main --htl 0 --healPercentage 0 get [EMAIL PROTECTED]
newfilename".

cvs -z3 update -dP -D "May 10"  -> success

cvs -z3 udpate -dP -D "May 20"  -> failure
13-Jun-03 6:35:34 PM (freenet.interfaces.LocalInterface, Interface # tcp/8481, ERROR): 
Unhandled throw accepting connections: Interface # tcp/8481
java.lang.NullPointerException
at freenet.transport.tcpConnection.(tcpConnection.java:123)
at freenet.transport.tcpListener.accept(tcpListener.java:48)
at freenet.interfaces.Interface.acceptConnections(Interface.java:217)
at freenet.interfaces.Interface.run(Interface.java:172)
at java.lang.Thread.run(Thread.java:334)

cvs -z3 update -dP -D "May 15"  -> success

cvs -z3 update -dP -D "May 17"  -> failure
13-Jun-03 6:55:00 PM (freenet.interfaces.LocalInterface, Interface # tcp/8481, ERROR): 
Unhandled throw accepting connections: Interface # tcp/8481
java.lang.NullPointerException
at freenet.transport.tcpConnection.(tcpConnection.java:123)
at freenet.transport.tcpListener.accept(tcpListener.java:48)
at freenet.interfaces.Interface.acceptConnections(Interface.java:217)
at freenet.interfaces.Interface.run(Interface.java:172)
at java.lang.Thread.run(Thread.java:334)

cvs -z3 update -dP -D "May 16"  -> no files changed from May 17, so didn't run

-fullversion for the broken May 17th build:
Engine: Just-in-time v3   Version: 1.1.x-cvs   Java Version: 1.1
Configuration/Compilation options:
  Compile date  : Fri Jun 13 18:51:45 EDT 2003
  Compile host  : pegasus
  Install prefix: /usr/local/kaffe
  Thread system : unix-jthreads
  CC: gcc
  CFLAGS: -g -O2 -Wall -Wstrict-prototypes
  LDFLAGS   : 
  ChangeLog head: 2003-05-15 Tim Stack <[EMAIL PROTECTED]>

-fullversion for the working May 15th build:
Engine: Just-in-time v3   Version: 1.1.x-cvs   Java Version: 1.1
Configuration/Compilation options:
  Compile date  : Fri Jun 13 18:41:35 EDT 2003
  Compile host  : pegasus
  Install prefix: /usr/local/kaffe
  Thread system : unix-jthreads
  CC: gcc
  CFLAGS: -g -O2 -Wall -Wstrict-prototypes
  LDFLAGS   : 
  ChangeLog head: 2003-05-11 Dalibor Topic <[EMAIL PROTECTED]>

CVS/Tag for the May 16th checkout which is the same as May 17th:
D2003.05.16.04.00.00

(I'm in US/Eastern a.k.a. EST5EDT.  I don't know what timezone
definition cvs is using when I say "May 16".)

After that, I ran cvs2cl.pl in the Kaffe source dir, which took an
*extremely* long time (easily longer than the entire testing process).
In fact, it's still not done yet, and it's been running for over
70 minutes.  So rather than wait for it, and attempt to get timestamps
on the individual check-ins from it, I'm just going to send this
message "as is".

-- 
Greg Wooledge  |   "Truth belongs to everybody."
[EMAIL PROTECTED]  |- The Red Hot Chili Peppers
http://wooledge.org/~greg/ |


pgp0.pgp
Description: PGP signature


Re: [kaffe] (sigh) Kaffe 1.1 can't do FEC decoding

2003-06-13 Thread Greg Wooledge
Dalibor Topic ([EMAIL PROTECTED]) wrote:

> yeah, the memory leaking is really weird. Have you tried running with
> --verbosemem to see where all that memory goes?

As you may recall, we tried to track it down a few months back by
applying a "counter" patch to kaffe, but it didn't work out because
Freenet never calls exit().  Then Mark J. Roberts (mjr) did something
which may or may not have been useful for you; it was a bit over
my head.

--verbosemem doesn't seem to produce any output with the early-May
CVS version.  I left the node running for about 10 minutes and
made a fair number of local requests (in addition to whatever came
in over the network).  I was seeing a lot of strange failures in
loading some of the images on some of the pages though; I didn't
attempt to track it down.

I repeated the experiment with 1.1.0 and within a few minutes it
had died:

$ java --verbosemem -mx 200M freenet.node.Main
Internal error: caught an unexpected exception.
Please check your CLASSPATH and your installation.
java/lang/OutOfMemoryError
Abort (core dumped) 

#0  0x401f9fcf in _thread_sys_kill ()
#1  0x401f98bb in abort ()
#2  0x400394e5 in unhandledException (eobj=0x3350d38) at exception.c:513
#3  0x40039387 in dispatchException (eobj=0x3350d38, baseframe=0x27f040)
at exception.c:470
#4  0x40038ffb in throwException (eobj=0x3350d38) at exception.c:235
#5  0x40038f77 in throwError (einfo=0x26f4b9c) at exception.c:217
#6  0x40056d0b in soft_fixup_trampoline (_data=0x3235cd) at soft.c:583
#7  0x4008280e in i386_do_fixup_trampoline () at machine.c:1187
#8  0x4005b4ae in callMethodV (meth=0x306560, func=0x3235c8, obj=0x26f4b9c, 
args=0x40056c76 "[\201Ãêó\003", ret=0x26f4ef8)
at ../../config/i386/common.h:38
#9  0x4005a5d3 in do_execute_java_method_v (obj=0x1588e00, 
method_name=0x4005c60d "finish", signature=0x4005becb "()V", mb=0x0, 
isStaticCall=0, argptr=0x26f4f6c "") at support.c:161
#10 0x4005a658 in do_execute_java_method (obj=0x1588e00, 
method_name=0x4005c60d "finish", signature=0x4005becb "()V", mb=0x0, 
isStaticCall=0) at support.c:187
#11 0x4005c6a0 in exitThread () at thread.c:441
#12 0x4005c5a7 in firstStartThread (arg=0x1588e00) at thread.c:402
#13 0x400857e8 in start_this_sucker_on_a_new_frame () at jthread.c:1266
#14 0x40085968 in jthread_create (pri=1073856528, func=0x97, 
daemon=1074116611, jlThread=0x10ff960, threadStackSize=0) at jthread.c:1336
#15 0x40096060 in _GLOBAL_OFFSET_TABLE_ ()
#16 0x8b000575 in ?? ()

-- 
Greg Wooledge  |   "Truth belongs to everybody."
[EMAIL PROTECTED]  |- The Red Hot Chili Peppers
http://wooledge.org/~greg/ |


pgp0.pgp
Description: PGP signature


Re: [kaffe] (sigh) Kaffe 1.1 can't do FEC decoding

2003-06-13 Thread Dalibor Topic
Hi Greg,

--- Greg Wooledge <[EMAIL PROTECTED]> wrote:
> So, after a week of failures every time I tried to download an FEC
> splitfile from Freenet, using Kaffe 1.1 (or the CVS version from
> right before the release), I decided that perhaps it's Kaffe that's
> causing the problem, rather than Freenet.

it usually is ;)

> Then I went back to the early-May Kaffe (and gmp-3.1.1) and the
> most recent Freenet build I have; I clicked it, took the defaults
> on the web form, and the operation finished in 37 seconds.  I had
> the file.

Hm, any chance of pinpointing the exact checkin on kaffe that started failing
FEC ? kaffe -fullversion of the last version that works, kaffe -fullversion of
the first version that fails, and an ugly binary search over checkouts, I guess
;) There have been quite a few changes to the core libraries in the last 4
weeks.


> freenet.node.ds.FSDataStoreElement$KeyInputStreamImpl.(FSDataSt
> oreElement.java:237)
> at

what does this code do?

> (Forgive me for not sewing the lines back together.)  I don't know
> if this error is directly related to the FEC splitfile errors; but
> on my system at least, Kaffe 1.1 fails to even *begin* an FEC decode,
> 100% consistently.  (And it leaks memory like an upside-down bucket.)

yeah, the memory leaking is really weird. Have you tried running with
--verbosemem to see where all that memory goes?

cheers,
dalibor topic

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

___
kaffe mailing list
[EMAIL PROTECTED]
http://kaffe.org/cgi-bin/mailman/listinfo/kaffe