-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Hello, all! :)
I've been working with Crypto++ for a while now, and have been keeping a
close eye on progress of 5.0. I downloaded it when it came out and compiled
it using gcc 3.2 without a hitch. Then, I ran the verification suite. A while
back when I was playing around with the 5.0 beta, I had noticed that
sometimes the second test of the blocking random number generator failed. I
was wondering if this still occurred under the 5.0 release. Thus, I gave it a
shot and tried to make it mess up.
I have discovered that I can make the test bomb predictably by just maxing
the CPU out at 100% while it's on the second test. For me that isn't too
hard. Just start 2-3 copies of Konquerer! :) Here's the output of the test
when it fails.
Using seed: 1033610324
Testing operating system provided blocking random number generator...
passed: it took 0 seconds to generate 16 bytes
FAILED: it generated 14588 bytes in 2 seconds
passed: 23166 generated bytes compressed to 23176 bytes by DEFLATE
Testing operating system provided nonblocking random number generator...
passed: 100000 generated bytes compressed to 100035 bytes by DEFLATE
Test ended at Wed Oct 2 21:58:57 2002
Seed used was: 1033610324
Just in case the seed was to blame I have tried the same test again using the
same seed, but letting the CPU usage remain at a normal level. This works
great! Here's the output of it working:
Using seed: 1033610324
Testing operating system provided blocking random number generator...
passed: it took 0 seconds to generate 16 bytes
passed: it generated 641 bytes in 15 seconds
passed: 1154 generated bytes compressed to 1159 bytes by DEFLATE
Testing operating system provided nonblocking random number generator...
passed: 100000 generated bytes compressed to 100035 bytes by DEFLATE
Test ended at Wed Oct 2 22:03:14 2002
Seed used was: 1033610324
Notice the HUGE difference in the number of bytes generated. Is this a bug in
Crypto++, or is there some weird condition that makes the operating system
random number generator go weird when the CPU maxes out? Either way, should I
be worried about this affecting the quality of the cryptography?
In case it helps any, my box has AMD K6-2 333MHz, and I'm running Gentoo
Linux. As I said before, I compiled with gcc 3.2. (However, I have noticed
this on the 5.0 beta with lesser versions of gcc as well.)
Hope this gets solved/fixed/figured-out!! If you need me to do any kind of
other tests to figure this out, just let me know! :)
Sincerely,
Hezekiah
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.7 (GNU/Linux)
iD8DBQE9m6bNicjSr4uiPQERAt6gAKCE7wAbtyaLmgL3pq6kIcmd99gZrACdH3A1
X4KxAKSxdEcVzF3LNgAoAB0=
=ngfD
-----END PGP SIGNATURE-----
