1 2015-01-05 00:00:18 <dzyk> Bitcointalk
 2 2015-01-05 00:00:22 <dzyk> Dzyk
 3 2015-01-05 00:00:32 <dzyk> Nvc inside
 4 2015-01-05 00:00:40 <dzyk> Dzyk.ru
 5 2015-01-05 00:00:42 <sipa> dzyk: ?
 6 2015-01-05 00:00:47 <dzyk> Da
 7 2015-01-05 00:01:21 <dzyk> Ya ne b
 8 2015-01-05 00:01:32 <sipa> english please
 9 2015-01-05 00:02:23 <dzyk> Please, ddos btce or i print 1 btc there
10 2015-01-05 00:02:31 <dzyk> Ili
11 2015-01-05 00:02:38 <sipa> not here
12 2015-01-05 00:02:43 <dzyk> Rasbante menya
13 2015-01-05 00:09:33 <user7779078> lol
14 2015-01-05 10:04:19 <jonasschnelli> sipa, while playing around with your logdb patch. I saw you create a logdb instance in main and changed all CWallet:: methods to use the this instance instead of the strWalletFile...
15 2015-01-05 10:05:06 <jonasschnelli> i would prefer to keep the strWalletFile in the constructor and maybe keep a static logdb instance in CWallet....
16 2015-01-05 10:06:46 <jonasschnelli> more clean interfaces (main/AppInit should only know CWallet and strWalletFile), less code diffs (more easy review)
17 2015-01-05 10:07:06 <jonasschnelli> s/more clean/cleaner :)
18 2015-01-05 10:15:38 <jgarzik> jonasschnelli, logdb is a fun thing to play with
19 2015-01-05 10:18:26 <jonasschnelli> jgarzik, could you see it/capable as new wallet db backend?
20 2015-01-05 10:19:30 <jgarzik> jonasschnelli, sure. That was the idea bandied about in IRC many moons ago.  The basic idea discussed was to replace BDB with an append-only mechanism.  Log-structured filesystems and databases provide plenty of background source material.
21 2015-01-05 10:22:35 <jonasschnelli> jgarzik, Okay. I still try to fit the pre-work from sipa as a wallet backend in current master. But i doubt if i should create a kind of switchable backend (by cmd arg) between bdb and logdb of if i should replace bdb completely.
22 2015-01-05 10:22:49 <jonasschnelli> I mean the migration tool could be written in python or so.
23 2015-01-05 10:23:23 <jonasschnelli> Support both backend would make things complicate... (more effort)
24 2015-01-05 10:23:26 <jgarzik> jonasschnelli, the goal is to drop BDB from bitcoind/Bitcoin-Qt.  A migration tool would be required, as you note.
25 2015-01-05 10:23:33 <jgarzik> no switching in bitcoind.
26 2015-01-05 10:24:15 <jgarzik> The --disable-wallet support was one step in that direction, as it successfully isolated BDB and proved it could be removed from the build.
27 2015-01-05 10:24:37 <jonasschnelli> jgarzik, i would see the migration as a detached process so bitcoind could be bdb free.
28 2015-01-05 10:25:25 <jonasschnelli> jgarzik, and IMO it's more sane to provide a detached process in migration so it's more conscious from the user perspective.
29 2015-01-05 10:25:35 <jgarzik> jonasschnelli, correct
30 2015-01-05 10:26:00 <jonasschnelli> okay. Then i might paddle in the right direction... let me continue then.
31 2015-01-05 10:44:14 <sipa> jonasschnelli: minimizing diffs is very nice
32 2015-01-05 15:26:50 <sipa> ;;blocks
33 2015-01-05 15:26:51 <altgribble> 337620
34 2015-01-05 22:52:16 <lclc> what would you use to send Bitcoin transactions from an embedded device? (source code of the app is c++)
35 2015-01-05 22:52:21 <lclc> or anyone knows a C++ SPV library?
36 2015-01-05 22:55:33 <gmaxwell> Eliel: ah 0.9.2 well there have been many memory usage improvements. But your number there is that its just using 672MB which is not that much.
37 2015-01-05 22:56:01 <Eliel> gmaxwell: also 500MB of swap.
38 2015-01-05 22:56:21 <gmaxwell> Eliel: The first memory column is _not_ swap, it's address space. And due to mmap, thread stacks, and other things there are large chunks of address space used which are never backed by memory.
39 2015-01-05 22:56:58 <phantomcircuit> wumpus, it does though since the entire wallet is basically loaded into memory and continuously checked when new blocks come in
40 2015-01-05 22:57:00 <Eliel> gmaxwell: there's nothing else running on that VPS and the swap had 500MB. Plus it was all gone after stopping bitcoind.
41 2015-01-05 22:57:15 <phantomcircuit> hell i think even when a new tx is accepted into the mempool they're all touched
42 2015-01-05 22:57:23 <phantomcircuit> so basically nothing in the wallet will get swapped out ever
43 2015-01-05 22:57:36 <Eliel> 19661 bitcoin   20   0  843m 615m 2236 S    3 80.6   2:46.77 bitcoind
44 2015-01-05 22:57:46 <Eliel> this is what it looks like after it's been fully restarted
45 2015-01-05 22:58:18 <Eliel> swap is at 185M
46 2015-01-05 22:58:42 <wumpus> phantomcircuit: I know, but that's nothing specific to the keypool, just to the number of keys in total in the wallet
47 2015-01-05 22:58:45 <gmaxwell> Eliel: yes sure? when you run more concurrent RPCs you'll see the address space usage go up a fair bit from additional rpc threads, that isn't using actual memory.
48 2015-01-05 22:59:07 <gmaxwell> Eliel: here is an example bitcoind on a busy 64 bit host which has no swap at all
49 2015-01-05 22:59:11 <gmaxwell> 21000 bitcoin+  20   0 2240568 873920  42580 S   6.1  5.3 484:45.49 bitcoind
50 2015-01-05 22:59:25 <Eliel> before:
51 2015-01-05 22:59:26 <Eliel> Swap:  1048568k total,   495932k used,   552636k free,    15312k cached
52 2015-01-05 22:59:38 <Eliel> after:
53 2015-01-05 22:59:39 <Eliel> Swap:  1048568k total,   188548k used,   860020k free,    61984k cached
54 2015-01-05 23:00:29 <Eliel> also, at one poin while bitcoind was still initializing: Swap:  1048568k total,    13828k used,  1034740k free,   162456k cached
55 2015-01-05 23:00:53 <sipa> Eliel: well if you're interested, please try 0.10.0rc1 and see whether things have improved
56 2015-01-05 23:01:06 <sipa> it's very unlikely we'll be making further memory usage improvements to the 0.9 line
57 2015-01-05 23:01:58 <gmaxwell> Eliel: ugh, you're also misreporting your swap in use.
58 2015-01-05 23:02:20 <gmaxwell> Eliel: please take a few minutes to research how linux swap usage is reported. memory that is cached is not in use.
59 2015-01-05 23:03:53 <Eliel> gmaxwell: I know how that works. but the cached amounts are negliglible.
60 2015-01-05 23:05:32 <gmaxwell> Eliel: How much memory does this host have?
61 2015-01-05 23:05:42 <Eliel> Mem:    781972k total,   769548k used,    12424k free,     1508k buffers
62 2015-01-05 23:06:02 <gmaxwell> Oh.
63 2015-01-05 23:06:33 <gmaxwell> Well that isn't enough especially not for 0.9.x and default configurations.
64 2015-01-05 23:07:21 <Eliel> it was running on 512MB up to one month ago :P
65 2015-01-05 23:07:50 <Eliel> the extra 256MB helped a lot. It was actually usable for a few weeks.
66 2015-01-05 23:07:52 <sipa> with swap i'm sure it would _work_
67 2015-01-05 23:08:00 <hearn> lclc: depends how tiny the device is
68 2015-01-05 23:08:00 <sipa> but with frequent thrashing
69 2015-01-05 23:08:14 <hearn> lclc: if it's really tiny, run a helper server and do as little as possible on device
70 2015-01-05 23:08:34 <Eliel> but if 0.10 is improving memory usage, I'll be happy to upgrade to that.
71 2015-01-05 23:08:47 <Eliel> but not before you consider it final as this is a production system.
72 2015-01-05 23:08:48 <hearn> lclc: if it's less tiny than that, and you want the device to talk p2p directly, you could use bitcoinj via jni. there is no c++ spv impl that i'm aware of, though i think jgarzik maintains a C impl
73 2015-01-05 23:16:14 <gmaxwell> Eliel: well you'll want to turn down the dbcache, as we increased it in 0.10, basically ofsetting the other memory usage improvements.
74 2015-01-05 23:17:26 <Eliel> I guess I'll go stick a dbcache=100 to the config file right now then :)
75 2015-01-05 23:18:18 <Eliel> ... or would 4 actually be a better idea for this server?
76 2015-01-05 23:20:10 <sipa> 100 is the default
77 2015-01-05 23:20:12 <sipa> in 0.10
78 2015-01-05 23:20:16 <sipa> in 0.9 it was 20 iirc
79 2015-01-05 23:24:39 <gmaxwell> If you're running out of ram, ... 4.
80 2015-01-05 23:25:00 <sipa> i doubt that matters
81 2015-01-05 23:25:18 <sipa> it still needs at least as much is needed to verify a full block
82 2015-01-05 23:55:55 <Eliel> bitcoind --help lists 100 as default
83 2015-01-05 23:56:11 <Eliel>   -dbcache=<n>           Set database cache size in megabytes (4 to 1024, default: 100)
84 2015-01-05 23:56:20 <sipa> ah, maybe it was raised in 0.9 alreayd
85 2015-01-05 23:56:46 <gmaxwell> ah I thought we raised it in 0.10. There is such a gap between changes and release that it can be hard to remember where things lined up