I've committed some changes to the way key comparison is handled
which should fix a math problem that led to an infinite loop and
a 100% CPU condition. Anyone who's been having this problem,
please update and let me know how things go.
Brave souls running unix nodes who wish to test Key.compare()
for accuracy, uncomment the debugging outputs in Key.compare().
Either redirect Fred's console output to a file and check it
using the attached scripts:
freenet_server >/tmp/f.out
bcparse /tmp/f.out | bccheck
Or, you could try something like this:
freenet_server | bcparse | nice -n 14 bccheck >check.out
In either case you are looking for any output from bccheck,
which will be a failure message with the offending numbers.
Be warned that the output of freenet_server with the debugging
outputs turned on can be huge. A busy node can easily acquire
tens or hundreds of megabytes of output in a few minutes. In
the first example you can watch the file size.. I'm not sure
what would happen in the second.
Attached are: Key.java with the debugging messages turned on,
bcparse (a Perl script), and bccheck (a gnu bc script; you need
gnu bc 1.06).
--
/* tavin cole
* composer of e-mail messages
*/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: comparetest.tar.gz
Type: application/octet-stream
Size: 3473 bytes
Desc: not available
URL:
<https://emu.freenetproject.org/pipermail/devl/attachments/20010221/6b303e24/attachment.obj>
