1 2013-06-19 02:34:07 <gmaxwell> http://labs.spotify.com/2013/06/18/creative-usernames/ I thought the bit at the end how subtle underlying library behavior submarined them was amusing.
2 2013-06-19 02:37:35 <gwillen> gmaxwell: thinking of database lock exhaustion? :-)
3 2013-06-19 02:37:43 <gwillen> (or whatever it was)
4 2013-06-19 02:44:29 <gmaxwell> Well, its a risk we face more generally than that. I liked that it was two layers deep here. I don't think I'd previously thought much about e.g. boost might depend on some underspecified libc or c compiler behavior.
5 2013-06-19 02:48:07 <Luke-Jr> I'm not sure I agree with the author's claim that it is correct to error if a character isn't part of Unicode 3.2 :p
6 2013-06-19 02:49:45 <gmaxwell> certantly the canonization function that couldn't handle the non-unicode 3.2 should have errored instead of screwed up.
7 2013-06-19 02:50:14 <gmaxwell> but at the end of the day the guy who exploited their vulnerable system didn't care if it was python or the library's fault.
8 2013-06-19 04:44:10 <tgs3> who was asking for site comparing all altcoins? gmaxwell or sipa? anyway http://dustcoin.com/mining and http://www.coinchoose.com/index.php
9 2013-06-19 04:45:06 <tgs3> oh it was jgarzik <@jgarzik> I wish there was an honest comparison written up somewhere
10 2013-06-19 05:49:12 <Luke-Jr> tgs3: he was asking for an *honest comparison*
11 2013-06-19 06:10:54 <coingenuity> anyone have any ideas why bitcoind might be running so poorly on an AMD APU?
12 2013-06-19 06:11:31 <Scrat> because you touch yourself
13 2013-06-19 06:11:41 <Scrat> cpu shouldn't matter
14 2013-06-19 06:11:49 <coingenuity> Scrat: shouldn't, but it seems to
15 2013-06-19 06:12:35 <TD> running so poorly when downloading the block chain?
16 2013-06-19 06:15:03 <Scrat> a typical APU is anemic compared to a desktop quadcore
17 2013-06-19 06:16:27 <gmaxwell> coingenuity: provide more data.
18 2013-06-19 06:17:04 <gmaxwell> Scrat: don't give bad advice; the cpu speed certantly matters in at least _some cases_, but coingenuity hasn't given enough data yet to know if he's talking about one of those. :)
19 2013-06-19 06:17:20 <gmaxwell> coingenuity: By running poorly do you mean that it spouts insults about your mother?
20 2013-06-19 06:17:23 <gmaxwell> :P
21 2013-06-19 06:17:57 <tgs3> Luke-Jr: what would be not honest, they are numbers. Any of them is falsified you think?
22 2013-06-19 06:18:31 <coingenuity> gmaxwell: well, basically, i've compared the ECDSA signing speed of my APU against a control amd64 system
23 2013-06-19 06:18:55 <coingenuity> the speed is around 1/4 that of the control, but blockchain sync is MUCH slower
24 2013-06-19 06:19:20 <coingenuity> IO makes a difference, of course, and the control has SSDs but it shouldnt cause such an extreme statistical disparity
25 2013-06-19 06:19:34 <coingenuity> so i'm wondering, what on earth could be causing bitcoind to operate SO slowly on an API
26 2013-06-19 06:19:36 <coingenuity> APU*
27 2013-06-19 06:20:14 <coingenuity> i'm watching it do initial-startup right now, and it's just passed block 100k on-disk after almost 2 minutes
28 2013-06-19 06:20:29 <coingenuity> so even verification of chain integrity is agonizingly bad
29 2013-06-19 06:20:39 <gmaxwell> woah
30 2013-06-19 06:20:43 <gmaxwell> yea, that sounds busted.
31 2013-06-19 06:20:47 <melvster> lakes me about a day to download a fresh client
32 2013-06-19 06:20:49 <coingenuity> extremely
33 2013-06-19 06:20:52 <Scrat> coingenuity: iowait%?
34 2013-06-19 06:21:23 <gmaxwell> melvster: pretty slow, but if you're talking about download that easily gets dominated by peer luck. But coingenuity is talking about something entirely different.
35 2013-06-19 06:21:33 <gmaxwell> multiminute startups don't make a lot of sense.
36 2013-06-19 06:21:40 <gmaxwell> coingenuity: what OS is on this thing?
37 2013-06-19 06:21:55 <coingenuity> Scrat: hm, io wait is negligable
38 2013-06-19 06:22:07 <coingenuity> gmaxwell: ubuntu 64
39 2013-06-19 06:22:10 <gmaxwell> With that kind of slowdown I wonder if some locking primitivte we're using misbehaves on that cpu.
40 2013-06-19 06:22:16 <gmaxwell> which cpu is it?
41 2013-06-19 06:22:24 <gmaxwell> gimme /proc/cpuinfo
42 2013-06-19 06:22:29 <coingenuity> AMD APU a3200 or something
43 2013-06-19 06:22:31 <coingenuity> sec
44 2013-06-19 06:22:48 <coingenuity> want a pastebin i assume?
45 2013-06-19 06:22:55 <gmaxwell> sure
46 2013-06-19 06:23:04 <coingenuity> kk, sec
47 2013-06-19 06:23:10 <coingenuity> its an e350 BTW
48 2013-06-19 06:24:18 <coingenuity> http://pastebin.com/0MqeEEXG here you are gmaxwell
49 2013-06-19 06:24:28 <coingenuity> thanks for taking a look :)
50 2013-06-19 06:26:29 <coingenuity> *sigh* only on block 140k
51 2013-06-19 06:26:31 <coingenuity> lmao
52 2013-06-19 06:26:41 <coingenuity> its not even syncing, just iterating through crap on disk
53 2013-06-19 06:27:08 <Scrat> it's a low voltage dual core :/
54 2013-06-19 06:27:30 <coingenuity> Scrat: it still shouldn't take 20 minutes to load the blockchain from disk
55 2013-06-19 06:28:09 <Scrat> is this a bootstrapped startup?
56 2013-06-19 06:28:16 <gmaxwell> ...
57 2013-06-19 06:28:27 <gmaxwell> coingenuity: Whats the storage on this thing?
58 2013-06-19 06:28:46 <gmaxwell> Any weird file system? e.g. ecryptfs?
59 2013-06-19 06:29:05 <coingenuity> no weird filesystem, brand new platter with ext3(if memory serves)
60 2013-06-19 06:30:08 <coingenuity> Scrat: its not bootstrapped, i just restarted the daemon to connect to a local peer to eliminate network-latency problems
61 2013-06-19 06:30:20 <coingenuity> and it's iterating through what was on disk already from the initial sync some months ago
62 2013-06-19 06:30:29 <gmaxwell> I'm not finding any redflags wrt that cpu.
63 2013-06-19 06:30:35 <gmaxwell> oh what bitcoin version is it?
64 2013-06-19 06:30:41 <coingenuity> 8.2
65 2013-06-19 06:30:47 <coingenuity> so should be real fast
66 2013-06-19 06:31:00 <coingenuity> its threading as it's supposed to, as well
67 2013-06-19 06:32:02 <coingenuity> in the time it's taken to go from startup ~> block 150k of what's on disk, my normal machines would be done doing a full sync already heh
68 2013-06-19 06:32:20 <gmaxwell> well, thats not a screaming cpu but I agree that it shouldn't be behaving like that.
69 2013-06-19 06:32:27 <gmaxwell> anything unusual showing up in the logs?
70 2013-06-19 06:33:13 <coingenuity> bitcoind's logs?
71 2013-06-19 06:33:23 <coingenuity> or syslog
72 2013-06-19 06:33:30 <gmaxwell> both, actually
73 2013-06-19 06:33:38 <gmaxwell> (well bitcoin's debug log and dmesg would be fine)
74 2013-06-19 06:33:46 <coingenuity> lets see
75 2013-06-19 06:34:04 <coingenuity> bwahaha
76 2013-06-19 06:34:14 <coingenuity> not really anything
77 2013-06-19 06:34:21 <coingenuity> just failed peerings
78 2013-06-19 06:34:31 <SomeoneWeird> well dont you fail
79 2013-06-19 06:34:33 <SomeoneWeird> :P
80 2013-06-19 06:35:36 <coingenuity> yeah, it's not showing me anything in debug.log
81 2013-06-19 08:11:35 <lucates> heh i won in some creepy slot machine 3.5 BTC
82 2013-06-19 09:25:01 <Subo1978> who has experience with BAMT and cgminer ?
83 2013-06-19 10:43:43 <jayhacks> i need bit coins
84 2013-06-19 10:44:19 <sipa> go buy them
85 2013-06-19 10:45:34 <coingenuity> lol
86 2013-06-19 11:20:45 <TD> sipa: i'll just squash all the commits now they're reviewed
87 2013-06-19 11:20:49 <TD> there's no point in having them be separate
88 2013-06-19 11:21:20 <sipa> TD: can you keep the move of CBlockLocator to .cpp separate?
89 2013-06-19 11:21:25 <sipa> TD: that feels like an unrelated change
90 2013-06-19 11:22:06 <TD> oh, sure
91 2013-06-19 11:22:10 <TD> of course
92 2013-06-19 11:34:52 <jgarzik> mornin'
93 2013-06-19 11:38:41 <phantomcircuit> jgarzik, so it is
94 2013-06-19 11:38:43 <phantomcircuit> :/
95 2013-06-19 12:20:29 <Luke-Jr> tgs3: well, it's not really a comparison; and I'd say its presentation implies they are more or less equals and the only thing that matters is profitability to miners
96 2013-06-19 12:25:33 <Vinnie_win> Where is the Jenkins server used to do the continuous integration for Bitcoin? Is it an Amazon cloud instance? Or does Jenkins provide servers for open source projects?
97 2013-06-19 12:28:05 <coingenuity> the heck is jenkins
98 2013-06-19 12:28:19 <TD> we don't use jenkins. there is a pull tester run by matt instead
99 2013-06-19 12:28:36 <Vinnie_win> oh hmm....was that written by hand or does it use a framework?
100 2013-06-19 12:31:27 <TD> was written by hand, i think
101 2013-06-19 12:31:45 <TD> sipa: can you summarise the pay-to-contract protocol talk?
102 2013-06-19 12:32:01 <jgarzik> ACTION is interested in P2C also
103 2013-06-19 12:42:01 <coingenuity> hm
104 2013-06-19 12:42:12 <coingenuity> anyone know if there's a bitcoind call to rebuild the block database?
105 2013-06-19 12:43:23 <sipa> coingenuity: -reindex
106 2013-06-19 12:43:32 <sipa> coingenuity: only at startup, not at runtime
107 2013-06-19 12:43:39 <buZz> -rescan Rescan the block chain for missing wallet transactions
108 2013-06-19 12:43:42 <buZz> not that?
109 2013-06-19 12:43:46 <coingenuity> ah, i found it just as you said so sipa :D
110 2013-06-19 12:43:50 <coingenuity> thank you much bud
111 2013-06-19 12:43:57 <sipa> buZz: no, that's for rescanning the chain for missing wallet transactions
112 2013-06-19 12:44:02 <coingenuity> ^
113 2013-06-19 12:44:02 <sipa> buZz: not for rebuilding the chain itself
114 2013-06-19 12:44:06 <buZz> oh right
115 2013-06-19 12:44:27 <coingenuity> hopefully it'll efficiently reindex the chain i have on disk so i dont have to move it over again lol
116 2013-06-19 12:44:41 <coingenuity> guessing its just a disparity in leveldb's checkpointing
117 2013-06-19 12:44:43 <sipa> it doesn't redownload
118 2013-06-19 12:45:00 <coingenuity> :)
119 2013-06-19 12:45:05 <coingenuity> eggcellent
120 2013-06-19 12:45:18 <coingenuity> been trying to get this 'thin' full node to play happily
121 2013-06-19 12:45:18 <epscy> coingenuity: what are you working on?
122 2013-06-19 12:45:29 <coingenuity> epscy: new bitcoind setup that uses far less power
123 2013-06-19 12:45:32 <coingenuity> AMD APU build
124 2013-06-19 12:45:32 <sipa> coingenuity: also, add -dbcache if you want to speed it up
125 2013-06-19 12:45:40 <coingenuity> sipa: oooooooo awesome
126 2013-06-19 12:45:43 <coingenuity> let me try that now
127 2013-06-19 12:45:46 <sipa> and have memory to spare
128 2013-06-19 12:45:50 <sipa> the argument is in MiB
129 2013-06-19 12:45:59 <coingenuity> i do, bitcoin is only using a few percent of memory for the moment
130 2013-06-19 12:46:10 <sipa> i usually set it to 1000 or so
131 2013-06-19 12:46:56 <coingenuity> sweet, got it running a reindex now
132 2013-06-19 12:47:04 <coingenuity> and thanks to dbcache it looks to be eating some more ram
133 2013-06-19 12:47:50 <coingenuity> oo nice
134 2013-06-19 12:47:54 <coingenuity> monitor is showing me blocks increasing now
135 2013-06-19 12:47:56 <coingenuity> very quickly too
136 2013-06-19 12:47:56 <tgs3> brother of Atkins
137 2013-06-19 12:48:00 <coingenuity> awesome sipa <3
138 2013-06-19 12:48:08 <coingenuity> i owe you about 10 beers now
139 2013-06-19 12:49:09 <Scrat> 30 mb wallet.dat
140 2013-06-19 12:49:26 <coingenuity> awesome
141 2013-06-19 12:49:29 <coingenuity> that reindex is /really/ fast
142 2013-06-19 12:49:32 <Scrat> when do I start experiencing slowness?
143 2013-06-19 12:49:32 <sipa> coingenuity: also, you can interrupt the reindex and continue later
144 2013-06-19 12:50:03 <coingenuity> sipa: ya, i figured since it's rebuilding it wont hurt to interrupt...but that's pretty sweet
145 2013-06-19 12:50:11 <coingenuity> it's indexing from .dat ~> leveldb ya?
146 2013-06-19 12:50:59 <sipa> coingenuity: correct, it rebuilds the block index based on blocks it finds in blocks/blk*.dat, and simultaneously connects those blocks (causing them to be verifying and rebuilding of the UTXO set)
147 2013-06-19 12:51:17 <coingenuity> ah, cool
148 2013-06-19 12:51:25 <coingenuity> thats kind of what i expected
149 2013-06-19 12:51:32 <coingenuity> (as i was moving from another node over)
150 2013-06-19 12:52:29 <coingenuity> that's awesome sipa - i am looking at it eating exactly 1000mb of ram in htop :D
151 2013-06-19 12:52:40 <epscy> coingenuity: what AMD APU do you have?
152 2013-06-19 12:52:40 <sipa> that's very accidental then
153 2013-06-19 12:52:46 <porquilho> coingenuity are you the developer of bitcoinity.org ?
154 2013-06-19 12:53:02 <coingenuity> epscy: its an e350 i think
155 2013-06-19 12:53:02 <sipa> the memory size is just using a heuristic (which is known to be off for larger values)
156 2013-06-19 12:53:10 <coingenuity> porquilho: no, just a friend of his, why?
157 2013-06-19 12:53:48 <Scrat> who here has dealt with massive wallets?
158 2013-06-19 12:54:01 <coingenuity> Scrat: not me!
159 2013-06-19 12:54:02 <sipa> Scrat: phantomcircuit
160 2013-06-19 12:54:03 <coingenuity> hehehe
161 2013-06-19 12:54:15 <porquilho> coingenuity just curious
162 2013-06-19 12:54:23 <sipa> Scrat: largest problem is a startup, afaik, with ~ >10000 keys
163 2013-06-19 12:54:39 <coingenuity> ^ has its issues
164 2013-06-19 12:54:43 <coingenuity> but also, sipa....
165 2013-06-19 12:54:51 <coingenuity> blockchain corruption is a big problem starting in 0.8
166 2013-06-19 12:55:03 <coingenuity> i have no clue what causes it, i think some malformed transactions
167 2013-06-19 12:55:04 <sipa> coingenuity: yes, and i don't understand why
168 2013-06-19 12:55:15 <sipa> coingenuity: my guess is crappy hardware, actually
169 2013-06-19 12:55:20 <phantomcircuit> sipa, wat
170 2013-06-19 12:55:23 <phantomcircuit> oh yeah
171 2013-06-19 12:55:26 <sipa> with 0.7 we saw more reports of people just getting stuck
172 2013-06-19 12:55:27 <phantomcircuit> Scrat, what about them
173 2013-06-19 12:55:31 <coingenuity> sipa: yeah, bad pages or hw could cause it
174 2013-06-19 12:55:39 <sipa> now it's more people complaining that their chainstate got corrupted
175 2013-06-19 12:55:43 <phantomcircuit> coingenuity, put that shit on ZFS
176 2013-06-19 12:55:45 <phantomcircuit> seriously
177 2013-06-19 12:55:48 <Luke-Jr> sipa: I think there's something more than just bad hw
178 2013-06-19 12:55:48 <Scrat> phantomcircuit: when does it start getting slow in normal operation?
179 2013-06-19 12:55:53 <sipa> Luke-Jr: maybe
180 2013-06-19 12:55:55 <coingenuity> ^
181 2013-06-19 12:55:58 <phantomcircuit> i dont trust bitcoind to operate correctly with a corrupt blockchain
182 2013-06-19 12:55:59 <sipa> Luke-Jr: but i have no better explanation yet
183 2013-06-19 12:56:06 <coingenuity> sipa: im running a BRAND NEW high end rig /just/ for bitcoind
184 2013-06-19 12:56:12 <coingenuity> i can't see anything bad in the hw there
185 2013-06-19 12:56:20 <sipa> and your chain got corrupted?
186 2013-06-19 12:56:26 <coingenuity> like, 10 times
187 2013-06-19 12:56:29 <sipa> wut
188 2013-06-19 12:56:32 <Luke-Jr> lol
189 2013-06-19 12:56:32 <phantomcircuit> Scrat, depends on what normal is for you, if you just have a lot of keys then really only the load time is an issue, if you have a lot of transactions then... everything is an issue
190 2013-06-19 12:56:33 <coingenuity> yeah...
191 2013-06-19 12:56:40 <Luke-Jr> coingenuity: what OS?
192 2013-06-19 12:56:40 <sipa> all reports i've heard were always on osx/windows
193 2013-06-19 12:56:46 <coingenuity> Luke-Jr: linux
194 2013-06-19 12:56:51 <Luke-Jr> sipa: mine was Linux
195 2013-06-19 12:56:52 <coingenuity> newest kernels and such
196 2013-06-19 12:56:53 <jgarzik> coingenuity, reproducible corruption? on what Linux version?
197 2013-06-19 12:56:59 <sipa> Luke-Jr: well, almost all
198 2013-06-19 12:57:08 <coingenuity> jgarzik: not sure if reproducable, i spoke with gmaxwell about it a while ago
199 2013-06-19 12:57:08 <Scrat> phantomcircuit: both, but I'm brute forcing it with hardware so far. and I don't know if 30 MB qualifies as big
200 2013-06-19 12:57:19 <coingenuity> jgarzik: 12.04LTS
201 2013-06-19 12:57:26 <phantomcircuit> Scrat, 30MB is tiny
202 2013-06-19 12:57:26 <sipa> coingenuity: bitcoind/bitcoin-qt?
203 2013-06-19 12:57:34 <coingenuity> either/or
204 2013-06-19 12:57:42 <sipa> running at shutdown?
205 2013-06-19 12:57:46 <coingenuity> nope
206 2013-06-19 12:57:47 <phantomcircuit> Scrat, make sure you're on 0.8.x and make sure you have ram for the entire blockchain
207 2013-06-19 12:57:59 <coingenuity> sipa: it pops up a chainstate-error out of the blue, then fails
208 2013-06-19 12:58:09 <coingenuity> sipa: but it doesnt segfault or cease operations, i.e. still responds to API calls
209 2013-06-19 12:58:19 <Scrat> phantomcircuit: ah right, a cached blockchain
210 2013-06-19 12:58:21 <coingenuity> bitcoind is better, it fails gracefully..
211 2013-06-19 12:58:22 <sipa> coingenuity: very interesting
212 2013-06-19 12:58:36 <coingenuity> sipa: yeah, quite troublesome
213 2013-06-19 12:58:38 <sipa> coingenuity: so not at startup, but during operation?
214 2013-06-19 12:58:40 <coingenuity> exactly
215 2013-06-19 12:58:48 <coingenuity> after days, weeks, or months of smooth operation
216 2013-06-19 12:58:49 <sipa> remarkable
217 2013-06-19 12:58:53 <coingenuity> *poop* it goes
218 2013-06-19 12:58:59 <coingenuity> yeah, its bizarre
219 2013-06-19 12:59:20 <coingenuity> i wish i'd held onto some of the debug logs i produced at various times during this process
220 2013-06-19 12:59:30 <coingenuity> next time it happens, i'll pastebin whatever i can dig up
221 2013-06-19 12:59:32 <sipa> i've never been able to reproduce a corruption myself, except by writing random crap to the leveldb files at runtime
222 2013-06-19 12:59:43 <sipa> and that's on a laptop that frequently overheats...
223 2013-06-19 12:59:56 <coingenuity> exactly - if this was crappy HW i could blame it on that
224 2013-06-19 13:00:07 <coingenuity> but it's a brand new machine, very high end, only a few months old
225 2013-06-19 13:00:08 <sipa> i know "works for me" isn't a good response, but i have no clue what causes it
226 2013-06-19 13:00:33 <coingenuity> absolutely, it's a strange phenominon(sp)
227 2013-06-19 13:00:35 <sipa> anything in particular you do with it?
228 2013-06-19 13:00:42 <sipa> specific RPC calls, for example
229 2013-06-19 13:02:35 <coingenuity> just the usual, getinfo type stuff, but in rather higher than usual volumes...can't imagine that'd cause any chainstate corruption
230 2013-06-19 13:02:54 <jgarzik> could create an RPC to call VerifyDB() during runtime
231 2013-06-19 13:02:54 <sipa> it shouldn't, but it's not my typical workload at least
232 2013-06-19 13:03:06 <sipa> jgarzik: ACK
233 2013-06-19 13:03:13 <jgarzik> getinfo has "interesting" locking issues
234 2013-06-19 13:03:41 <coingenuity> when i poked around a little it /seems/ to my untrained eyes that it has something to do with gigantic and possibly malformed transactions in the in/output history of something in my wallet
235 2013-06-19 13:03:46 <jgarzik> pulls info from a bunch of unrelated information domains
236 2013-06-19 13:03:46 <sipa> how so? it grabs cs_main and cs_wallet, just like everything else
237 2013-06-19 13:03:58 <sipa> coingenuity: wait, what error do you exactly get?