--- Day changed Sun Jul 22 2018
02:15 < waxwing> alexcato see my response
02:16 < waxwing> if i'm wrong something is completely crazy, we addressed this years ago ... but i'm sorry i had to spend several hours concentrating on the previous bug in restarting with subset yesterday, and i'll have to do the same today (testing my fixes in great detail), and i'm busy all next week.
02:16 < waxwing> so really have to concentrate on that to get the bugfix out.
02:17 < waxwing> oh in sweep? yes i think, or at least seem to recall, that is a known issue. well, at least i remember it, i don't know how "known" it is.
02:18 < waxwing> #256 in joinmarket may not have discussed that specific aspect of how sweep is a bit screwy. maybe it can be fixed, i can't remember now, but like i say, have to concentrate on the other thing.
02:27 < AgoraRelay> [agora-irc/AlexCato] hi waxwing, not expecting you to fix this right away; its actually not *that* hard to fix, so i might even try myself.
02:27 < AgoraRelay> [agora-irc/AlexCato] Basically the only change needed is to have only 1 offer per maker in the offerslist (the cheapest one), before passing that to the offer selection algorithm
02:29 < AgoraRelay> [agora-irc/AlexCato] python does not seem to make this easy though, the checking for dictionary value duplicates in just one of the fields
02:30 < AgoraRelay> [agora-irc/AlexCato] also, not only sweep affected. The weighted algorithm has just made it a less severe problem, but still gives an unfair advantage to makers with many overlapping offers
02:31 < waxwing> alexcato, but see my response. that's what that block of code does.
02:31 < waxwing> your example was sweep, right
02:31 < AgoraRelay> [agora-irc/AlexCato] hm. Reading your answer on github, gimme some time to understand that
02:31 < AgoraRelay> [agora-irc/AlexCato] if that already restricts it to the cheapest one, the fix for sweep is trivial
02:31 < waxwing> it's a twisty bit of code. but of course this was a huge deal, we spent time making sure of that.
02:32 < waxwing> re: sweep it may or may not be trivial. have you read #256? if it is indeed trivial, then great (although in that case it was a bad oversight from back in the day)
02:33 < waxwing> and of course i could be wrong about non-sweep, in which case it's another huge screwup. but i don't think that code even changed (you can probably find it identical in the old codebase)
02:34 < AgoraRelay> [agora-irc/AlexCato] which #256? Not that high in joinmarket-clientserver yet, in joinmarket that issue seems to be something else: https://github.com/JoinMarket-Org/joinmarket/issues/256
02:35 < waxwing> yes i referred to it above, i guess you didn't see
02:35 < waxwing> oh mistake sorry
02:35 < AgoraRelay> [agora-irc/AlexCato] ah, 356
02:35 < waxwing> 356     
02:36 < AgoraRelay> [agora-irc/AlexCato] will have a look and re-test. I might very well have overlooked something
02:36 < waxwing> ah yes now scanning it again, it's connected, it's exactly the issue you're noticing
02:36 < waxwing> i saw it as a quandary with no obvious resolution, but it's far from obvious until you think about the reasoning explained there
02:40 < AgoraRelay> [agora-irc/AlexCato] i think back in the day, nearly all makers had many different offers. Nowadays, the default is 1 maker, 1 offer. Only very few makers change that. So using the full amount of the mixdepth to be sweeped as the cjamount will be a good approximation; basically, it can just work as the non-sweep selection
02:41 < AgoraRelay> [agora-irc/AlexCato] because as a last step, it filters out offers which do not fit any more after calculating fees
02:41 < AgoraRelay> [agora-irc/AlexCato] if a maker has just 1 offer, this does neither advantage/disadvantage anyone, because if it gets filtered out in the last step, it couldnt have be a participant in the CJ anyways
02:42 < waxwing> yes there's probably an improvement along those lines
02:42 < waxwing> i just want you to be sure to understand how it works now, and why :) and that there isn't a bug like that in non-sweep right?
02:43 < AgoraRelay> [agora-irc/AlexCato] unsure how much cjfees were paid to makers back in the day. Maybe then it was a big factor? If so, that increases the need for the current sweep algo
02:44 < AgoraRelay> [agora-irc/AlexCato] yeah, understood why it is the way it is
02:44 < AgoraRelay> [agora-irc/AlexCato] about this:
02:44 < AgoraRelay> [agora-irc/AlexCato]  and that there isn't a bug like that in non-sweep right? <-- unsure yet, not coming from python, the code snippet you quoted i dont understand what it does yet. Need to up my skills here first
02:45 < waxwing> it's an unusual structure. it took me a while to get it. but it should do what the comment says.
02:46 < AgoraRelay> [agora-irc/AlexCato] i'll just go the safe route here, set a breakpoint and just check that the dict indeed has no duplicate makers at this point
02:47 < waxwing> actually i understand now what the confusion is.
02:47 < waxwing> it's because there's deduplication code lower down, after selection.
02:48 < waxwing> as i recall that's left there because of the manual choice option
02:48 < AgoraRelay> [agora-irc/AlexCato] yeah, that tripped me up. But it's clear to me now why it's there, even though not needed usually
02:48 < waxwing> if you use -P and choose an offer for counterparty X, all its other offers get removed on the next go-round
03:00 < AgoraRelay> [agora-irc/AlexCato] experimentally confirmed: not a problem with non-sweeps
04:19 -!- GitHub71 [GitHub71@gateway/service/github.com/x-ylljschfrfqiwdbu] has joined #joinmarket     
04:19 < GitHub71> [joinmarket-clientserver] AlexCato opened pull request #172: Fixes the sweep maker selection issue (#171) by moving to approximate cjamount estimation (master...fix_sweep_makerselection) https://git.io/fNlLX
04:19 -!- GitHub71 [GitHub71@gateway/service/github.com/x-ylljschfrfqiwdbu] has left #joinmarket []     
07:15 -!- GitHub148 [GitHub148@gateway/service/github.com/x-ypimwhkxbrtjskbp] has joined #joinmarket     
07:15 < GitHub148> [joinmarket-clientserver] AdamISZ opened pull request #173: Fixes bugs in restart-with-subset including #169; see PR for more (master...restart-with-subset-2) https://git.io/fNlOf
07:15 -!- GitHub148 [GitHub148@gateway/service/github.com/x-ypimwhkxbrtjskbp] has left #joinmarket []     
07:41 -!- GitHub143 [GitHub143@gateway/service/github.com/x-rsclwzegrjdhbezh] has joined #joinmarket     
07:41 < GitHub143> [joinmarket-clientserver] AdamISZ pushed 2 new commits to master: https://git.io/fNl3G
07:41 < GitHub143> joinmarket-clientserver/master 7df69d0 fivepiece: add the wallets directory to gitignore
07:41 < GitHub143> joinmarket-clientserver/master 57ada0c AdamISZ: Merge #170: add the wallets directory to gitignore...
07:41 -!- GitHub143 [GitHub143@gateway/service/github.com/x-rsclwzegrjdhbezh] has left #joinmarket []     
07:41 -!- GitHub152 [GitHub152@gateway/service/github.com/x-migqfnroxmobftmz] has joined #joinmarket     
07:41 < GitHub152> [joinmarket-clientserver] AdamISZ closed pull request #170: add the wallets directory to gitignore (master...gitignore_add_wallets) https://git.io/fNWX3
07:41 -!- GitHub152 [GitHub152@gateway/service/github.com/x-migqfnroxmobftmz] has left #joinmarket []     
07:42 -!- GitHub174 [GitHub174@gateway/service/github.com/x-zxxyugcmghfuzqjm] has joined #joinmarket     
07:42 < GitHub174> [joinmarket-clientserver] AdamISZ pushed 3 new commits to master: https://git.io/fNl3l
07:42 < GitHub174> joinmarket-clientserver/master 62b4fe6 fivepiece: escape regex for '.'
07:42 < GitHub174> joinmarket-clientserver/master be79a06 fivepiece: manual install for joinmarket-qt
07:42 < GitHub174> joinmarket-clientserver/master a0ee299 AdamISZ: Merge #168: update manual joinmarket-qt install steps in readme...
07:42 -!- GitHub174 [GitHub174@gateway/service/github.com/x-zxxyugcmghfuzqjm] has left #joinmarket []     
07:42 -!- GitHub138 [GitHub138@gateway/service/github.com/x-hajxpnbnbzrmunht] has joined #joinmarket     
07:42 < GitHub138> [joinmarket-clientserver] AdamISZ closed pull request #168: update manual joinmarket-qt install steps in readme (master...jmqt_symlink_venv) https://git.io/fNWz8
07:42 -!- GitHub138 [GitHub138@gateway/service/github.com/x-hajxpnbnbzrmunht] has left #joinmarket []     
07:45 -!- GitHub58 [GitHub58@gateway/service/github.com/x-tyyzdpzqiqczlilv] has joined #joinmarket     
07:45 < GitHub58> [joinmarket-clientserver] AdamISZ pushed 1 new commit to master: https://git.io/fNl3g
07:45 < GitHub58> joinmarket-clientserver/master 4a15cbd AdamISZ: Merge #173: Fixes bugs in restart-with-subset including #169; see PR for more...
07:45 -!- GitHub58 [GitHub58@gateway/service/github.com/x-tyyzdpzqiqczlilv] has left #joinmarket []     
07:45 -!- GitHub184 [GitHub184@gateway/service/github.com/x-thoywnohesgrdtmr] has joined #joinmarket     
07:45 < GitHub184> [joinmarket-clientserver] AdamISZ closed pull request #173: Fixes bugs in restart-with-subset including #169; see PR for more (master...restart-with-subset-2) https://git.io/fNlOf
07:45 -!- GitHub184 [GitHub184@gateway/service/github.com/x-thoywnohesgrdtmr] has left #joinmarket []     
09:28 -!- lnostdal [~lnostdal@77.70.119.51] has quit [Quit: https://www.Quanto.ga/]     
09:51 < arubi> might be of interest, I think it's new : https://travis-ci.org/fivepiece/joinmarket-clientserver/jobs/406857098#L3601
10:28 -!- lnostdal [~lnostdal@84-238-146-90.ip.btc-net.bg] has joined #joinmarket     
11:14 -!- lnostdal [~lnostdal@84-238-146-90.ip.btc-net.bg] has quit [Ping timeout: 240 seconds]     
11:28 < waxwing> (1) couldn't reproduce locally after repeated tries (2) there is randomness in tweak_tumble_schedule so theoretically we could have a nondeterministic failure there but (3)it's very weird if such a thing led to that kind of index error. sorry bit exhausted won't be checking it further today.
11:31 < arubi> sure sure, I'm just posting for reference
11:31 -!- qubenix [~qubenix@185.232.22.99] has joined #joinmarket     
11:35 < waxwing> sorry i don't mean to come across tetchy, that was supposed to be informative :)
11:35 < arubi> I understand 100% :)
11:36 -!- lnostdal [~lnostdal@77.70.119.51] has joined #joinmarket     
11:36 < arubi> working on adding caching dependencies to install.sh and travis so we don't encounter these curl in the future so I'm going through a lot of builds
11:37 < arubi> (so these less common errors are coming up)
11:37 < arubi> er, these curl errors in the future*
12:20 -!- undeath [~undeath@unaffiliated/undeath] has joined #joinmarket     
14:26 -!- qubenix [~qubenix@185.232.22.99] has quit [Remote host closed the connection]     
14:28 -!- GitHub142 [GitHub142@gateway/service/github.com/x-iybejxgfjfguumzz] has joined #joinmarket     
14:28 < GitHub142> [joinmarket-clientserver] AlexCato opened pull request #174: Switch default merge_algorithm in joinmarket.cfg to 'greedy' (master...switch_mergealgo_default) https://git.io/fNlEk
14:28 -!- GitHub142 [GitHub142@gateway/service/github.com/x-iybejxgfjfguumzz] has left #joinmarket []     
14:34 -!- undeath [~undeath@unaffiliated/undeath] has quit [Quit: WeeChat 2.0.1]     
16:27 -!- qubenix [~qubenix@66.55.147.60] has joined #joinmarket     
16:33 -!- qubenix [~qubenix@66.55.147.60] has quit [Quit: qubenix]     
16:33 -!- qubenix [~qubenix@66.55.147.60] has joined #joinmarket     
17:16 -!- qubenix [~qubenix@66.55.147.60] has quit [Remote host closed the connection]     
17:17 -!- AgoraRelay [~jmrelayfn@p5DE4AFE6.dip0.t-ipconnect.de] has quit [Ping timeout: 244 seconds]     
17:26 -!- puddinpop [~puddinpop@unaffiliated/puddinpop] has quit [Ping timeout: 264 seconds]     
17:27 -!- M1 [~Michail@michail.com] has quit [Ping timeout: 244 seconds]     
17:30 -!- M1 [~Michail@michail.com] has joined #joinmarket     
17:32 -!- AgoraRelay [~jmrelayfn@p5DE4AB3D.dip0.t-ipconnect.de] has joined #joinmarket     
17:39 -!- M1 [~Michail@michail.com] has quit [Ping timeout: 240 seconds]     
17:41 -!- M1 [~Michail@michail.com] has joined #joinmarket     
19:15 -!- puddinpop [~puddinpop@unaffiliated/puddinpop] has joined #joinmarket     
21:28 -!- d3spwn [~jimmy@a83-161-157-200.adsl.xs4all.nl] has joined #joinmarket     
22:25 < d3spwn> does joinmarket support the new rpc cookie auth?
23:26 < waxwing> d3spwn, yes; see the release notes
23:26 < waxwing> unless it's changed or something?
23:26 < waxwing> assume no one has a problem with the switch to greedy as default? it only ends up grabbing a lot of space temporarily, in equilibrium i think it's just better.
23:38 < waxwing> might make more sense to change the name of 'default' to something else though, if it's no longer the default :)
23:45 < d3spwn> there dont seem to be any makers on testnet :(
23:59 < waxwing> no; haven't run any myself since last year some time