Thus spake Karsten Loesing (karsten.loesing@xxxxxxx): > These are just minor issues, as you can see. I think they are good suggestions. I'll be fixing them soon. > I think the most promising approaches to find more bugs would be: 1) > write some tests, 2) run the code in a private Tor network, 3) ask > more people to review your code, and/or 4) just put it in 0.2.2.x > and hope to find the bugs before 0.2.2.x becomes the new stable. 1) > and 2) are probably rather painful. If you want to do 2) and need > support, please let me know. Yeah, I am inclined to try to avoid #2 if possible, as I'm not sure how much that actually buys us for the time investment. So long as existing clients can properly handle the new consensus documents without dying, I think it is better if I can devote my time to testing how our current authority platforms fare in actually computing integer arithmetic the same, and in writing statistical tests using my 'EXTENDCIRCUIT 0\n' control port patch in my other branch to ensure that the weights are being used by clients as we expect for the current network. I have a feeling that setting up a test tor network large and diverse enough to perform these tests in a worthwhile fashion will consume about another week of development time at least, possibly for the both of us, and if we're not careful about how we do it, it won't tell us anything new anyways. But maybe I'm just impatient :) Right now my directory authority is producing votes with the new code, and it is producing fake consensus v9 documents for purposes of running the verification code in that branch against them. So far the consensus is parseable, and the weights being generated do satisfy the balancing equations from the other thread. -- Mike Perry Mad Computer Scientist fscked.org evil labs
Attachment:
pgpZEZLer5Eov.pgp
Description: PGP signature