On Sun, Apr 11, 2010 at 05:20:07PM -0700, Jamie Zawinski wrote: > I believe this was fixed in xscreensaver 5.06 in July 2008. > > The problem was a single character typo in yarandom.c: it said "024" > (octal) where it should have said "24" (decimal), causing the two lags > to not be co-prime.
lakeview ok % diff -u unicode-screensaver-0.2/utils/yarandom.c xscreensaver-5.10/utils/yarandom.c; echo $? 0 So, no, this is not fixed. The issue that was present in the yarandom.c I patched was that the value of seed when the generator is being initialized becomes identical after around i == 7. My conjecture is that you're basically losing randomness since seed is multiplied by a large number. This pushes whatever randomness existed from the additional seed to higher and higher bits until it disappears altogether. By using a bitwise rotate, we move that randomness to a variety of bits and therefore the avalanche effect is dramatically increased. Files dump[4-6] in #573989 show how the generator produces identical numbers very quickly, and using diff on pairs of these files shows how the seed variable ends up as a constant. The issue is not the actual generator at all, but the seeding. Whether the algorithm varies slightly from Knuth seems to be mostly irrelevant here. -- brian m. carlson / brian with sandals: Houston, Texas, US +1 832 623 2791 | http://www.crustytoothpaste.net/~bmc | My opinion only OpenPGP: RSA v4 4096b: 88AC E9B2 9196 305B A994 7552 F1BA 225C 0223 B187
signature.asc
Description: Digital signature
