1 2013-02-26 20:24:13 <MC1984> hmm bitcoin wallet initial sync is still pretty slow here
2 2013-02-26 20:24:26 <MC1984> from a peer with protocol: 70001
3 2013-02-26 20:32:34 <MC1984> hm i wouldnt mind seeing netowrd send/reciee rates for the peers in the peer list
4 2013-02-26 20:33:18 <MC1984> also maybe data transfer per session/total in the options, lots of bandwidth sensitive mobile apps have that
5 2013-02-26 20:33:20 <MC1984> important information
6 2013-02-26 20:34:28 <andytoshi> has the freenode hidden service been down for days for anyone else?
7 2013-02-26 20:35:26 <gavinandresen> MC1984: "patches welcome"
8 2013-02-26 20:36:00 <sipa> gavinandresen: i assume he's talking about the android app
9 2013-02-26 20:36:11 <MC1984> yes
10 2013-02-26 20:36:34 <MC1984> my keyboard is really bad damn
11 2013-02-26 20:36:50 <gavinandresen> sipa: open source android app?
12 2013-02-26 20:37:04 <sipa> gavinandresen: yes; Goonie's
13 2013-02-26 20:37:15 <gavinandresen> mmm. I bet he'd welcome patches, too
14 2013-02-26 20:37:53 <MC1984> LOL PATCHES WELCOME MOTHERFUCKER
15 2013-02-26 20:38:02 <MC1984> im not demanding, just saying what i think
16 2013-02-26 20:38:09 <gavinandresen> okey dokey.
17 2013-02-26 20:38:33 <sipa> \\o/ 3.2% extra signature verification speedup
18 2013-02-26 20:38:45 <Goonie> MC1984: android has data usage monitoring built in, no need to replicate that into the apps
19 2013-02-26 20:38:49 <sipa> (doing benchmark again, as it may just be statistical variation)
20 2013-02-26 20:39:17 <MC1984> oh yeah youre right, ive seen that in the setting somewhere
21 2013-02-26 20:39:20 <sipa> and before gavinandresen complains: it's just one line of course code: EC_GROUP_precompute_mults(group, ctx);
22 2013-02-26 20:39:54 <Goonie> MC1884: The reason its still slow is because of slow disc access. Bloom filtering is currently more for saving bandwidth.
23 2013-02-26 20:40:28 <gavinandresen> I dunno??? another line of code?????? lets spend a hour or two talking about it first.
24 2013-02-26 20:40:28 <MC1984> thats ok, not really complaining
25 2013-02-26 20:41:09 <gavinandresen> for starters, shouldn't it be 'context' instead of 'ctx' ? Or maybe ptrCtx ? Or ptrContext?
26 2013-02-26 20:41:27 <sipa> haha
27 2013-02-26 20:41:29 <gavinandresen> (sorry, feeling silly today)
28 2013-02-26 20:41:45 <Luke-Jr> p_ctx_ecdsa_group_forsigcheck
29 2013-02-26 20:42:04 <Goonie> MC1984: and if you're talking about "initial startup" or about "replay up to the birth date of your wallet", when actually bloom filtering is not even used.
30 2013-02-26 20:42:04 <sipa> pctxGroupMultsPrecompute
31 2013-02-26 20:42:24 <Luke-Jr> pctxGroupMultsPrecomputeOrange
32 2013-02-26 20:42:37 <Goonie> s/when/then
33 2013-02-26 20:42:38 <sipa> pctxGroupMultsPrecomputeOrangeEleven
34 2013-02-26 20:42:45 <sipa> ^- pretty sure gavin will prefer this
35 2013-02-26 20:42:49 <MC1984> i think td explained this to me a few days ago actually lol
36 2013-02-26 20:42:52 <Luke-Jr> pctxGroupMultsPrecomputeOrangeElevenIsHuInTonal
37 2013-02-26 20:43:18 <phantomcircuit> psh
38 2013-02-26 20:43:28 <phantomcircuit> just call it i
39 2013-02-26 20:43:36 <Luke-Jr> but i is for integers!
40 2013-02-26 20:43:39 <MC1984> hmm what is satoshi client v 0.7.99 though
41 2013-02-26 20:43:40 <phantomcircuit> any good programmer should know what it is from context
42 2013-02-26 20:43:53 <Luke-Jr> MC1984: pre-0.8
43 2013-02-26 20:44:06 <MC1984> oh
44 2013-02-26 20:44:19 <MC1984> its the protocol version that mattes right
45 2013-02-26 20:45:00 <Luke-Jr> depends what you're looking at it for
46 2013-02-26 20:45:28 <sipa> phantomcircuit: even when the variable itself is (a pointer to) the context? :p
47 2013-02-26 20:45:43 <sipa> yes yes i know, different context
48 2013-02-26 20:45:51 <gmaxwell> sipa: 3.2% is small compared to Hal though!
49 2013-02-26 20:46:42 <sipa> gmaxwell: it is on top of hal, though
50 2013-02-26 20:46:45 <gmaxwell> oh!
51 2013-02-26 20:47:03 <sipa> i would have expected a larger effect, actually
52 2013-02-26 20:47:53 <sipa> i remember i tried this before, but couldn't see an effect and therefor assumed the openssl code already did the precomputation
53 2013-02-26 20:48:14 <sipa> however, there is absolutely no call to that function anywhere in openssl's source tree
54 2013-02-26 20:48:21 <phantomcircuit> sipa, oh well in that case it should clearly be p for pointer
55 2013-02-26 20:48:47 <phantomcircuit> sipa, what function?
56 2013-02-26 20:49:18 <phantomcircuit> there are actually an inordinate number of things in openssl that are generated through macros and dont appear to be used at all
57 2013-02-26 20:49:22 <phantomcircuit> or defined anywhere
58 2013-02-26 20:49:30 <sipa> EC_GROUP_precompute_mults
59 2013-02-26 20:49:32 <phantomcircuit> like pretty much all of the ASN.1 functions
60 2013-02-26 20:50:54 <phantomcircuit> sipa, not only is it not called but it appears to not even exist
61 2013-02-26 20:50:56 <phantomcircuit> macros
62 2013-02-26 20:51:07 <sipa> eh, i may have made a type
63 2013-02-26 20:51:29 <phantomcircuit> nope that's the right name
64 2013-02-26 20:51:30 <sipa> int EC_GROUP_precompute_mult(EC_GROUP *group, BN_CTX *ctx);
65 2013-02-26 20:51:41 <phantomcircuit> oh right the s
66 2013-02-26 20:52:11 <phantomcircuit> ./crypto/ec/ec_key.c:\treturn EC_GROUP_precompute_mult(key->group, ctx);
67 2013-02-26 20:53:01 <phantomcircuit> which is EC_KEY_precompute_mult
68 2013-02-26 20:53:21 <phantomcircuit> which isn't called from anywhere but speed.c
69 2013-02-26 20:53:22 <sipa> yeah, i just searched for calls to anything 'precompute'
70 2013-02-26 20:53:29 <gavinandresen> gmaxwell sipa: FYI : I bought the domains "bitcoincore.com/.org" today, and am planning on buying a cheap SSL certificate for one or both of them; I need a SSL cert to start generating and testing "real" payment requests.
71 2013-02-26 20:54:09 <sipa> cool
72 2013-02-26 20:55:46 <kinlo> gavinandresen: if you want a really cheap cert, startssl.com provides them for free
73 2013-02-26 20:55:53 <kinlo> in case you didn't know that yet :)
74 2013-02-26 20:56:52 <sipa> gavinandresen: i'm a big fan of the payment protocol idea, but i just don't have the time to be very actively involved in it now
75 2013-02-26 20:56:56 <phantomcircuit> im guessing he wants one that will actually work everywhere
76 2013-02-26 20:57:00 <gavinandresen> kinlo: I'm going to splurge and spend $25 for a good-for-5-years cert
77 2013-02-26 20:57:02 <phantomcircuit> :|
78 2013-02-26 20:57:20 <kinlo> gavinandresen: as you wish, just trying to help :)
79 2013-02-26 20:57:42 <gavinandresen> sipa: RE: payment protocol: no problem, most of the work is Qt stuff
80 2013-02-26 20:57:42 <gmaxwell> phantomcircuit: the startssl certs work everywhere except on old browsers, but this is for the payment protocol stuff in any case.
81 2013-02-26 20:57:55 <phantomcircuit> gmaxwell, i know from experience that isn't true
82 2013-02-26 20:58:07 <phantomcircuit> startssl screwed up a while ago
83 2013-02-26 20:58:19 <phantomcircuit> the certs they issue are in a chain 4 certificates long
84 2013-02-26 20:58:39 <phantomcircuit> root -> intermediary -> intermediary -> server cert
85 2013-02-26 20:58:51 <phantomcircuit> various browsers have the root/first intermediary
86 2013-02-26 20:58:55 <gavinandresen> ooh, good test case. I may have to get one of those, too
87 2013-02-26 20:59:00 <phantomcircuit> so it's impossible to be sure which to provide
88 2013-02-26 20:59:05 <gmaxwell> phantomcircuit: sure, you have to supply the chain.
89 2013-02-26 20:59:15 <phantomcircuit> gmaxwell, yes but
90 2013-02-26 20:59:22 <kinlo> phantomcircuit: the standard doesn't limit the depth of the chain, so startssl is not to blame imho
91 2013-02-26 20:59:28 <sipa> seems my benchmark (reindex the chain until block 160k, with -nocheckpoints) has very little variation actually: 1340.56s vs 1339.33s
92 2013-02-26 20:59:29 <gmaxwell> But thats not just a startssl thing, current verisign certs have a four cert long chain.
93 2013-02-26 20:59:36 <phantomcircuit> if the browser is webkit with the first intermediary and you supply the root ca also
94 2013-02-26 20:59:36 <sipa> so it seems the 3.2% is quite accurate
95 2013-02-26 20:59:42 <phantomcircuit> it wont validate
96 2013-02-26 20:59:50 <phantomcircuit> even thought there is a valid trust paht
97 2013-02-26 20:59:56 <kinlo> phantomcircuit: so a webkit bug ? :)
98 2013-02-26 20:59:57 <gmaxwell> sipa: we should probably have an ECDSA unit test that can double as a microbenchmark. it would save you some time.
99 2013-02-26 21:00:01 <phantomcircuit> so you're stuck between two bad choices
100 2013-02-26 21:00:12 <kinlo> oh well, certificates are bad by design
101 2013-02-26 21:00:36 <sipa> so it seems to reduce the (CPU) time from 1385s to 1340s
102 2013-02-26 21:00:43 <sipa> nothing spectacular, but hey :)
103 2013-02-26 21:00:45 <gmaxwell> (and might save our bacon, e.g. if openssl started accepting a public key of 0 as valid or whatever)
104 2013-02-26 21:00:51 <phantomcircuit> gmaxwell, i've dealt with this a few times now and only figured out exactly the issue relatively recently, it's quite annoying
105 2013-02-26 21:06:29 <MC1984> every little helps sipa
106 2013-02-26 21:13:45 <andytoshi> has anyone else noticed that the freenode hidden service is down?
107 2013-02-26 21:14:12 <Luke-Jr> ACTION wonders if anyone here would even care :p
108 2013-02-26 21:16:02 <andytoshi> it is a bit silly :P you have to auth to freenode to use it
109 2013-02-26 21:16:28 <andytoshi> then it says "andytoshi/tor-sasl" when you connect
110 2013-02-26 21:16:48 <andytoshi> but it decreases the signal-to-noise ratio on the tor network, so i do it as a public service
111 2013-02-26 21:20:34 <sipa> sounds like an excuse for doing an IBD over tor
112 2013-02-26 21:21:15 <andytoshi> sipa: i did do one, actually
113 2013-02-26 21:21:32 <andytoshi> and i'm running a hidden service bitcoin node, so hopefully i'll return the favor..
114 2013-02-26 21:51:28 <MC1984> do some android apps install themselves to extrnal SD card by default?
115 2013-02-26 21:51:38 <MC1984> like bitcoin wallet, and i know i didnt put it there
116 2013-02-26 21:53:11 <MC1984> in fact im still not sure how android names its volumes, i think the system partition is "internal storage" and the otehr internal volume is called an "SD card" ?
117 2013-02-26 21:55:38 <HM> it's arbitrary
118 2013-02-26 21:55:49 <HM> for phones that don't have sdcard slots /sdcard is internal also
119 2013-02-26 21:56:14 <HM> i recall my G1 used to put apps on SD card automatically, may have been a cyanogenmod feature
120 2013-02-26 22:06:07 <MC1984> i have mnt/sdcard which is internal and mnt/external_sd
121 2013-02-26 22:07:06 <MC1984> of sdcard, 1GB is "app storage" (???) rest is generally available
122 2013-02-26 22:07:45 <doublec> my note 2 doesn't even allow storing apps on the sd card
123 2013-02-26 22:07:54 <MC1984> and i have a 32gb external SD which i was asspaind to find i couldnt just install everything on it
124 2013-02-26 22:08:18 <doublec> same. I have a 64gb external card but can only store apps on the tiny internal partition.
125 2013-02-26 22:08:18 <MC1984> so now its mainly full of playstation roms
126 2013-02-26 22:09:09 <MC1984> yeah they chose 1gb out of 8gb for "app storage" on this device
127 2013-02-26 22:09:16 <MC1984> its rooted so there must be a way
128 2013-02-26 22:11:45 <phantomcircuit> MC1984, it depends a lot on the firmware you have loaded
129 2013-02-26 22:12:13 <HM> ACTION looks at Linux's in-kernel keyring and wonders why nobody uses it
130 2013-02-26 22:12:26 <phantomcircuit> cm typically lets you install apps to sdcard and even move already installed apps
131 2013-02-26 22:12:36 <phantomcircuit> although on more recent phones you should have plenty of space
132 2013-02-26 22:13:01 <MC1984> no cyanogen for this dvice yet if ever
133 2013-02-26 22:16:03 <MC1984> they dont seem to give a shit about anything that doesnt say Nexus, Samsung, LG, Sony, etc on it
134 2013-02-26 22:18:58 <HM> ACTION strokes his Nexus
135 2013-02-26 22:19:04 <MC1984> and manufacturers dont seem to give a shit about supporting released devices so much as releasing new ones
136 2013-02-26 22:19:19 <sipa> MC1984: why would that be :p
137 2013-02-26 22:19:28 <MC1984> in fact from what ive seen, theyre using different versions of android to justify selling 200 different devices
138 2013-02-26 22:19:52 <MC1984> zero incentive to update me to jelly bean!
139 2013-02-26 22:20:21 <phantomcircuit> MC1984, what phone do you have o.O
140 2013-02-26 22:20:37 <HM> i think it's that there's 0 incentive to update existing phones
141 2013-02-26 22:20:45 <HM> user just blame app authors or google
142 2013-02-26 22:20:48 <HM> users*
143 2013-02-26 22:20:51 <MC1984> its an archos cobalt 80
144 2013-02-26 22:20:52 <MC1984> tablet
145 2013-02-26 22:21:04 <phantomcircuit> oh
146 2013-02-26 22:21:06 <MC1984> sold new with ICS 2 months ago
147 2013-02-26 22:21:10 <phantomcircuit> yeah not gonna happen
148 2013-02-26 22:21:14 <MC1984> but its was only 50 quid
149 2013-02-26 22:21:31 <MC1984> and it has one of the strongest chipsets out there i gather
150 2013-02-26 22:21:42 <sipa> powah!
151 2013-02-26 22:21:49 <MC1984> lol
152 2013-02-26 22:22:31 <MC1984> its really nice for the money
153 2013-02-26 22:22:53 <MC1984> i just went on their site looking for firmware updates and instead i found INTRODUCING THE ARCHOS PLATINUM RANGE NOW WITH JELLY BEAN
154 2013-02-26 22:22:58 <MC1984> and i realised how this works
155 2013-02-26 22:23:39 <phantomcircuit> MC1984, it's got similar specs to my phone but with a larger screen
156 2013-02-26 22:23:49 <phantomcircuit> also i seriously doubt they're going to update it
157 2013-02-26 22:24:06 <MC1984> yeh
158 2013-02-26 22:24:20 <phantomcircuit> at 50 gbp per device they wont have the profit margin to support the device
159 2013-02-26 22:24:23 <phantomcircuit> basically at all
160 2013-02-26 22:24:34 <MC1984> well it sells for 120 now
161 2013-02-26 22:24:38 <MC1984> i got a sale
162 2013-02-26 22:24:48 <MC1984> bought 3 of the fuckers for family
163 2013-02-26 22:24:56 <MC1984> android for everyone
164 2013-02-26 22:26:23 <Goonie> MC1984: since android 4, internal and external is on the same filesystem for most phones, so it doesn't really matter
165 2013-02-26 22:27:21 <MC1984> ye
166 2013-02-26 22:28:08 <muhoo> archos does stuff weird tho, or at least they used to
167 2013-02-26 22:28:30 <muhoo> IIRC they are still shipping a tablet with donut on it.
168 2013-02-26 22:28:46 <MC1984> they used to make a lot of rubbish i think
169 2013-02-26 22:29:02 <MC1984> but since andorid they sem to have some decent products
170 2013-02-26 22:29:12 <MC1984> the gamepad thing looks like a great psp alternative
171 2013-02-26 22:29:12 <muhoo> ok, bitcoind 0.8 question. is there a way to get it to listen, when it is still connected to just one node? this is for some testing i want to do
172 2013-02-26 22:29:24 <sipa> muhoo: yes, use -listen
173 2013-02-26 22:29:29 <muhoo> thx
174 2013-02-26 22:29:52 <muhoo> is listen=yes in the .conf file equivalent?
175 2013-02-26 22:30:02 <sipa> listen=1
176 2013-02-26 22:30:08 <sipa> =yes may also work, unsure
177 2013-02-26 22:30:36 <BlueMatt> how did one exploit the bug that was because duplicate entries in the merkle tree were equivalent to the padded merkle tree?
178 2013-02-26 22:30:43 <muhoo> awesome, listen=1 does the job, thanks
179 2013-02-26 22:31:00 <gmaxwell> BlueMatt: you make a node reject a block by feeding it the second preimage block first.
180 2013-02-26 22:31:07 <gmaxwell> then it will never accept the real block.
181 2013-02-26 22:31:08 <BlueMatt> ahh, ok thanks
182 2013-02-26 22:31:28 <sipa> BlueMatt: take incoming block, duplicate some of the last transactions to make it a power of two, relay it; anyone who accepts it will consider the real blockchain invalid
183 2013-02-26 22:31:50 <sipa> gmaxwell beats me
184 2013-02-26 22:32:08 <gmaxwell> thats what you get for giving a more complete answer.
185 2013-02-26 22:36:13 <muhoo> wow, bitcoinj can go stir-crazy if it can't reach a peer, 150% CPU.
186 2013-02-26 22:36:15 <muhoo> 2013-02-26 15:34:51,056 INFO com.google.bitcoin.core.PeerGroup: Peer discovery took 0msec
187 2013-02-26 22:36:45 <MC1984> where do you get that log
188 2013-02-26 22:37:40 <muhoo> MC1984: https://www.refheap.com/paste/11855
189 2013-02-26 22:38:19 <muhoo> something racy maybe
190 2013-02-26 22:38:32 <MC1984> yeah but how did you get it out of your android
191 2013-02-26 22:38:39 <muhoo> i'm not running on an android :-)
192 2013-02-26 22:38:59 <muhoo> i'm developing in clojure on linux, using bitcoinj as my bitcoin backend
193 2013-02-26 22:39:09 <MC1984> oh
194 2013-02-26 22:39:29 <sipa> java is used on more than just android :p
195 2013-02-26 22:39:50 <MC1984> yes i know im just in android mode or something lol
196 2013-02-26 22:40:03 <muhoo> and the jvm is used for more than just java (thankfully!)
197 2013-02-26 22:41:32 <muhoo> whoa... connections: 117
198 2013-02-26 22:50:24 <TD> muhoo: it's not a race
199 2013-02-26 22:50:28 <TD> just dumb peer management logic
200 2013-02-26 22:50:36 <Goonie> muhoo: yes, I filed that a while ago
201 2013-02-26 22:50:48 <TD> when a peer disconnects or fails to connect it says "i need a new peer", and because there's only one it gets the same one back
202 2013-02-26 22:50:52 <TD> tries to connect, fails, infinite loop
203 2013-02-26 22:51:02 <TD> there's no delay between trying to reconnect
204 2013-02-26 22:51:18 <TD> the code assumes it has a peer discovery object, really
205 2013-02-26 22:51:35 <sipa> just a 0.1s sleep before trying to connect would do miracles, perhaps
206 2013-02-26 22:51:59 <TD> ACTION shrugs
207 2013-02-26 22:52:07 <TD> there are rules about when you can and cannot sleep
208 2013-02-26 22:52:13 <TD> it's not a high priority to fix
209 2013-02-26 22:52:18 <TD> there are more important bugs, unfortunately
210 2013-02-26 22:52:56 <TD> i won't come back to another round of network improvements for another version or two
211 2013-02-26 22:55:24 <muhoo> https://www.refheap.com/paste/11856
212 2013-02-26 22:55:50 <muhoo> yep, no big deal, might add that myself
213 2013-02-26 22:57:35 <TD> sleep is the wrong solution
214 2013-02-26 22:57:41 <muhoo> fyi it also does that if the peer disconnects for any reason
215 2013-02-26 22:57:45 <TD> the real issue is that it's retrying peers after they fail
216 2013-02-26 22:57:47 <TD> yes
217 2013-02-26 22:58:06 <TD> it should try peers and if they disconnect, move on to the next. the issue then being what if the network dies
218 2013-02-26 22:58:18 <TD> so possibly there should be a single retry per peer, or something
219 2013-02-26 22:58:21 <TD> it needs thought
220 2013-02-26 22:58:22 <muhoo> thundering herd problem
221 2013-02-26 22:58:34 <TD> that's a lot of peers
222 2013-02-26 22:58:38 <TD> how many did you go up to?
223 2013-02-26 22:59:08 <muhoo> i'm not sure it was peers and not botched connection attempts from rpc. not sure what that connections count actualy means
224 2013-02-26 22:59:49 <Goonie> the problem is also it connects to the same peer multiple times, if no more addresses can be discovered
225 2013-02-26 23:00:00 <muhoo> this particular bitcoind is ONLY connecting (with connect=ipaddress) to one node!
226 2013-02-26 23:00:45 <muhoo> and, linux was running out of inodes after bitcoinj started going ape-shit and reconnecting
227 2013-02-26 23:00:49 <TD> haha
228 2013-02-26 23:01:03 <TD> ACTION -> sleep
229 2013-02-26 23:01:08 <muhoo> indeed.
230 2013-02-26 23:01:48 <muhoo> not going to yak shave too many weirdnessess with what is just a dummy test network anyway
231 2013-02-26 23:18:57 <ProfMac> anyone know about FIX?