On Thu, Jan 10, 2002 at 07:44:17PM -0800, Ian Clarke wrote: > I have set up a Freenet node running with the new datastore code at: > > signature=d3a28bffcc707e5e0b998bcdce9917b43f1c5c8b,d8905f79a1fdd2a1461a4ef5d8a56a5ecb052165 > version=Fred,0.4,1.45,450 > presentations=1 > identity.g=930168de21e7fb66c0375e08e964255a0f7f0ad54507a51864afdc686f36be8bb8b7865408116060c5f34f94b5146cbef9e4adb70324fba01d34c1c60817cbadf6854d654176cb391de0d41e0f0fbbc8ceea5546c09a676b0d9a9988c7a1ce36ce31596037a18b4d540374bdf2ad071a3f8dd1015a9d8ba0f0d51cde212db6da > identity.q=ef1f7a7a73362e526515f348075aee265e9eff45 > identity.p=cb0a782c7abff4920000000000000000000000000000000000000000023d662854a10e52de49da383d9ee21d7a337213d24ed096f95a5d37b8537bbaa58a2a6b26bd328f6a32cec77180f78d5be43d80e813e4018d09da38bd58fd615c01fbab492ec203c69e3da9fd682ce8aa98f15ad8057970edb44fe1ed08e0462e5b8d97 > identity.y=7d968525a2846cd0acf8bdb277f88175b38e5220a22e9eff0ba0799295ffd28737f345b332bd3b4047681faae74d7b499e1569b7bd42f76e073cdd20912c74d09fb49aab98f604341e6de1e09a058a850ebb52f64381b4a666df49621fd545bb4ecdc4b13e6f905953b3b18df81139eac04bac33b00703d6ce6bae662592503a > sessions=1 > > Tavin wants to try to keep the networks reasonably separate for the > moment, so anyone trying it out (everyone REALLY SHOULD try it out!) > please just connect to this node and don't "pollute" it with any of the > old unreliable nodes. > > I doubt this cleanliness will last too long but what the hell.
Well, that's easy, change the last good build number to 450. I didn't do that so that new_datastore nodes could join the existing network without forcing everyone to upgrade, since the new_datastore code is largely untested so far. > The new datastore code makes heavy use of 64bit numbers, which seem to > cause real problems for IBM's JRE when JIT compiling is enabled (which > is the default). Also, jikes 1.15 has trouble compiling it (IBM's > fault, not our's), so jikes users should use 1.14. Kaffe users should > use the latest CVS (although I couldn't get it working at all so I am > using the latest blackdown JRE). No heavier use than the old datastore, or any part of the node that deals with freenet keys since we have always assumed they could be longer than 2^30. > You should log on "minor" or "debug" if you have a big hard-disk, and > keep a close-eye on it for problems. There is one problem that I am aware of: when retrieving a key via FCP you may get garbage data the first time you ask for it. This garbage may start at the beginning, or start occurring partway through good data. In any event the next time you request that data (once it's cached in the store) there is no trace of the garbage. I suspect some kind of braino in the low-level lock handling code that should keep read streams blocking behind the stream writing the data. Working on it, but the keyboard in my laptop decided to die yesterday.. -- :: tavin cole (tcole at espnow.com) :: _______________________________________________ Devl mailing list Devl at freenetproject.org http://lists.freenetproject.org/mailman/listinfo/devl
