1 2013-09-11 00:11:28 <warren> gmaxwell: the latest 2979 doesn't have -O by default
 2 2013-09-11 00:12:50 <warren> cfields: ^
 3 2013-09-11 00:12:52 <gmaxwell> warren: I assumed that would be another pull, cfields linked a patch.
 4 2013-09-11 00:12:57 <warren> oh
 5 2013-09-11 00:14:17 <cfields> gmaxwell: want that one now? it's just the one-liner
 6 2013-09-11 00:15:13 <warren> might as well, nobody objects to that
 7 2013-09-11 00:15:39 <gmaxwell> cfields: sure.
 8 2013-09-11 00:21:22 <cfields> gmaxwell: hmm, benchmark results are a bit too promising. anything i need to be aware of as far as cache poisoning?
 9 2013-09-11 00:21:50 <gmaxwell> cfields: how too promising? and which did you check first?
10 2013-09-11 00:22:30 <gmaxwell> the total amount of data read is only 8gb or so, so cache shouldn't matter that greatly.
11 2013-09-11 00:22:43 <cfields> gmaxwell: i just let reindex run for a while, picked a random range's timestamps, then repeated with the rebuilt leveldb
12 2013-09-11 00:23:17 <gmaxwell> okay, it matters greatly which blocks you're looking at, so you need to make sure you compare with the same heights.
13 2013-09-11 00:23:52 <cfields> sure, i compared the same height ranges
14 2013-09-11 00:24:15 <cfields> 149155-174095, to be specific
15 2013-09-11 00:25:36 <gmaxwell> sounds good, what kind of improvement?
16 2013-09-11 00:25:50 <gmaxwell> I wouldn't be shocked if it were large, we're sort of surprisingly slow right now.
17 2013-09-11 00:27:13 <cfields> time was 6:38 vs 2:45
18 2013-09-11 00:29:08 <gmaxwell> awesome. Cool.
19 2013-09-11 00:29:17 <gmaxwell> Sounds like we'll have some node sync world records soon.
20 2013-09-11 00:29:41 <cfields> anything you'd like me to do with that so it doesn't get forgotten?
21 2013-09-11 00:48:43 <Luke-Jr> yay, pulltester is happy
22 2013-09-11 00:49:58 <Luke-Jr> gmaxwell: ^
23 2013-09-11 00:50:05 <warren> cfields: PR it?
24 2013-09-11 00:50:20 <Luke-Jr> warren: it's already part of pt2
25 2013-09-11 00:50:34 <warren> URL?
26 2013-09-11 00:50:50 <Luke-Jr> it's my autoconf_pt2 branch - to be PR'd when pt1 is merged
27 2013-09-11 00:51:25 <Luke-Jr> pt1 just takes care of the immediate build issues
28 2013-09-11 00:51:36 <Luke-Jr> afaik lack of -O isn't messing up the build for anyone?
29 2013-09-11 00:51:47 <gmaxwell> merged.
30 2013-09-11 00:52:29 <warren> Luke-Jr: I look forward to seeing pt2
31 2013-09-11 00:53:26 <Luke-Jr> pt2 PR'd: https://github.com/bitcoin/bitcoin/pull/2986
32 2013-09-11 00:54:12 <warren> Luke-Jr: I thought you said cfields's DNDEBUG for leveldb build was in this.
33 2013-09-11 00:54:33 <Luke-Jr> warren: no, I thought we were talking about the -O1 for debug/default builds
34 2013-09-11 00:54:58 <cfields> i'm still playing with the NDEBUG thing, looks like that might've indeed been too good to be true
35 2013-09-11 00:57:40 <cfields> Luke-Jr: we decided to use -02 for now, to match the old build. then it's up for debate for permanant change
36 2013-09-11 01:06:40 <rdymac> "Bitcoin is a worldwide community base from businessmen to anarchists, rich to poor and someday maybe even females will start to use it. "
37 2013-09-11 01:20:42 <cfields> Luke-Jr: you're kinda shotgunning through a bunch of changes I had planned to address, each with its own discussion...
38 2013-09-11 01:21:27 <warren> cfields: what do you mean by "too good to be true"?
39 2013-09-11 01:23:26 <cfields> warren: with proper optims, that discrepancy seems to go away
40 2013-09-11 01:24:01 <Luke-Jr> cfields: I didn't figure any of these changes would be controversial
41 2013-09-11 01:24:28 <Luke-Jr> rdymac: my wife uses bitcoins
42 2013-09-11 01:25:02 <rdymac> it was a quote from a response in the forum, I found it funny enough to be shared here
43 2013-09-11 01:25:45 <rdymac> :)
44 2013-09-11 02:40:37 <hydromet> hello, has anyone used safello in Sweden? https://safello.com/sv/index.xhtml
45 2013-09-11 02:42:33 <hydromet> btw, this article < http://www.arcticstartup.com/2013/09/04/swedish-bitcoin-startup-safello-emphasises-transparency > interviewing the Safello CEO includes a mention of "theory" that the NSA is in fact behind Bitcoin intentionally.
46 2013-09-11 02:42:41 <hydromet> Could Satoshi be the NSA?
47 2013-09-11 02:43:02 <hydromet> After the Snowden crucible, people's imaginations run wild!
48 2013-09-11 02:52:21 <Luke-Jr> cfields: any idea what pulltester is doing here to break this? :/
49 2013-09-11 03:01:52 <Diablo-D3> hydromet: hes just repeating stuff Ive said for years
50 2013-09-11 03:02:11 <Diablo-D3> that the CIA probably invented bitcoin to make it easier to do black ops undetected in difficult situations
51 2013-09-11 03:46:02 <cfields> Luke-Jr: where?
52 2013-09-11 03:47:42 <cfields> Luke-Jr: that leveldb change needs a bunch of fixes, but i have to get some work done before I can play anymore
53 2013-09-11 03:48:01 <Luke-Jr> cfields: it works fine for me no matter how I build it..
54 2013-09-11 03:52:00 <cfields> Luke-Jr: leveldb is sensitive to build/srcdir changes because automake descends into it blindly
55 2013-09-11 03:52:18 <cfields> there's no need to subsitute the leveldb.a target when using the system one, it'll just be happily ignored
56 2013-09-11 03:52:48 <cfields> instead, just add -lleveldb and -lmemenv as needed to $LIBS
57 2013-09-11 03:53:24 <Luke-Jr> well, can't be linking to leveldb.a too
58 2013-09-11 03:53:37 <cfields> it won't build it, ofc
59 2013-09-11 03:54:44 <cfields> Luke-Jr: just change it to use -lleveldb -lmemenv, and set the -L paths appropriately
60 2013-09-11 03:55:30 <Luke-Jr> IIRC that doesn't work properly with automake - it wants the .a form when using a .a it's supposed to build
61 2013-09-11 03:56:00 <cfields> it's not building it though, it's blind to it
62 2013-09-11 03:56:40 <Luke-Jr> it's supposed to.
63 2013-09-11 03:57:25 <cfields> i meant, automake isn't actively building leveldb. it hands off to a 3rd party makefile, then regains control when it's complete
64 2013-09-11 03:57:44 <cfields> so it's not like a normal noinst static
65 2013-09-11 03:57:49 <Luke-Jr> right, but from the perspective of the automake project, it thinks it is
66 2013-09-11 03:58:12 <Luke-Jr> it sees leveldb.a as the target, and executes commands to build it
67 2013-09-11 03:58:36 <cfields> that's just plain make, that's not an automake target
68 2013-09-11 03:59:49 <Luke-Jr> I'm not aware of a distinction in this case.
69 2013-09-11 04:00:37 <cfields> in this case, bitcoind is an automake target. so bitcoind_LDADD and friends are special. automake resolves dependencies for them
70 2013-09-11 04:01:38 <Luke-Jr> but it won't see -lleveldb in bitcoind_LDADD and depend on leveldb.a - it will ignore it unless we use the .a name directly
71 2013-09-11 04:05:53 <cfields> blarg, looking at it, i suppose you're right on that one. I remembered it looking a bit different
72 2013-09-11 04:13:22 <cfields> Luke-Jr: for the libs, use something like this: http://pastebin.com/raw.php?i=4KA0j6Eh
73 2013-09-11 04:13:45 <cfields> where LEVELDB_LIBS is subst'd from configure, and INTERNAL_LEVELDB is an am conditional
74 2013-09-11 04:14:37 <cfields> if using internal, LEVELDB_LIBS will be empty and the conditional will be true, so the local targets will be added and built
75 2013-09-11 04:15:20 <Luke-Jr> I don't see how that'd be different than what it's doing right now.. but I'm falling asleep, so I'd better look again tomorrow
76 2013-09-11 04:15:24 <cfields> otherwise, '-L/foo/bar -lleveldb -lmemenv' should be coming from configure
77 2013-09-11 04:19:25 <cfields> Luke-Jr: i haven't looked into the includes changes yet, but my point is that you don't need to change the leveldb targets at all
78 2013-09-11 04:20:14 <Luke-Jr> cfields: IIRC I didn't until it broke the build unless I did
79 2013-09-11 04:20:28 <Luke-Jr> should make be matching up the different paths that refer to the same location?
80 2013-09-11 04:20:41 <cfields> no
81 2013-09-11 04:21:03 <cfields> foo/ != foo/../foo/