1 2014-04-29 00:01:53 <dexX7> i see the blockhash of the orphaned block and 0 confirmations for those transactions
  2 2014-04-29 00:02:54 <dexX7> just to make sure, we are on the same level: i'm talking about searchrawtransactions which accesses the address index
  3 2014-04-29 00:06:08 <sipa> and was the transaction confirmed in another branch?
  4 2014-04-29 00:07:23 <dexX7> yes
  5 2014-04-29 00:33:03 <dexX7> alright. i figured it out: actually both transactions are listed - the confirmed one and the orphaned one
  6 2014-04-29 00:42:26 <dexX7> (and.. that's what you were saying a few lines earlier - so ty :)
  7 2014-04-29 02:41:22 <jgarzik> Ubuntu upgrade 13->14 suddenly decided I did not need boost anymore.
  8 2014-04-29 02:41:30 <jgarzik> lovely
  9 2014-04-29 02:42:30 <shesek> wow, from not having any decent testnet blockexplorer a few months ago, we have 4 now - webbtc, blockr, insight and biteasy
 10 2014-04-29 02:42:51 <shesek> and also blockexplorer that had testnet for a long time, but its somewhat broken :\
 11 2014-04-29 05:26:07 <justanotheruser> Should transactions in the mempool be cleared every week or so (or maybe the time can be determined based on transaction amount)? When we have floating fees, I assume more than a few transactions will need to be rebroadcasted with higher feea
 12 2014-04-29 05:29:54 <Luke-Jr> justanotheruser: why?
 13 2014-04-29 05:30:30 <justanotheruser> Luke-Jr: because they will pay a fee too low
 14 2014-04-29 05:30:37 <Luke-Jr> justanotheruser: so just allow them to be replaced?
 15 2014-04-29 05:30:45 <justanotheruser> Luke
 16 2014-04-29 05:31:04 <justanotheruser> Luke-Jr: why not?
 17 2014-04-29 05:31:20 <michagogo> cloud|justanotheruser: he's saying, have replace-by-fee
 18 2014-04-29 05:31:25 <Luke-Jr> because maybe they have sufficient fee and will be mined, then you have to redownload them\
 19 2014-04-29 05:31:30 <michagogo> cloud|(I think)
 20 2014-04-29 05:31:58 <justanotheruser> Luke-Jr: are you suggesting they push to pool if their tx fails?
 21 2014-04-29 05:31:58 <Luke-Jr> michagogo|cloud: there are lots of possible replacement algorithms
 22 2014-04-29 05:32:10 <Luke-Jr> justanotheruser: no? that doesn't even make sense
 23 2014-04-29 05:32:32 <Luke-Jr> justanotheruser: there's just zero value to removing transactions from the memorypool based on time
 24 2014-04-29 05:33:03 <Luke-Jr> if you're trying to save memory, then you don't care how long they've been in there (although you might bias toward removing older ones)
 25 2014-04-29 05:33:07 <justanotheruser> Luke-Jr: okay, well at least allow someone to spend the same inputs after a little time
 26 2014-04-29 05:33:12 <Luke-Jr> exactly
 27 2014-04-29 05:33:48 <Luke-Jr> justanotheruser: the code for *that* already exists; it's just not enabled by the *default* policy
 28 2014-04-29 05:33:49 <justanotheruser> Okay, so this is something that may be implemented?
 29 2014-04-29 05:33:54 <justanotheruser> Oh, I see
 30 2014-04-29 05:33:55 <petertodd> Luke-Jr: speaking of, did you see my post on how you don't need cpfp for replace-by-fee?
 31 2014-04-29 05:34:28 <gmaxwell> Luke-Jr: hey now, you're going a bit too far— since the mempool is currently in memory forgetting them can decrease memory usage. :)
 32 2014-04-29 05:34:29 <Luke-Jr> petertodd: I saw it, I understand it, but instead of CPFP you need some kind of upgraded payment protocol..
 33 2014-04-29 05:34:40 <Luke-Jr> gmaxwell: [05:33:03] <Luke-Jr> if you're trying to save memory, then you don't care how long they've been in there (although you might bias toward removing older ones)
 34 2014-04-29 05:34:47 <petertodd> Luke-Jr: or simply SIGHASH_ANYONECANPAY
 35 2014-04-29 05:36:29 <Luke-Jr> it's too bad Linux doesn't have some kind of "weak pages" a stdlib could use to implement smarter caching.
 36 2014-04-29 05:36:39 <Cray-on-> does someone know php?
 37 2014-04-29 05:36:39 <Luke-Jr> (memory pages the kernel is free to drop at will)
 38 2014-04-29 05:36:51 <Cray-on-> got a quick question
 39 2014-04-29 05:36:56 <petertodd> Luke-Jr: you sure it doesn't? I think I've seen that before
 40 2014-04-29 05:37:07 <Luke-Jr> petertodd: I've never heard of such a thing before, at least
 41 2014-04-29 05:37:51 <gmaxwell> Luke-Jr: https://lwn.net/Articles/594847/
 42 2014-04-29 05:38:20 <Luke-Jr> realistically, it'd probably be better to have some way for the kernel to express "please free up some memory, we're out of spare caching space" so programs can do their own prioritisation algo
 43 2014-04-29 05:39:23 <Luke-Jr> gmaxwell: hm!
 44 2014-04-29 05:39:38 <Luke-Jr> 
ACTION wonders if anything actually uses this

 45 2014-04-29 05:43:38 <gmaxwell> (see both MADV_FREE and MADV_DONTNEED)
 46 2014-04-29 05:57:08 <wumpus> Luke-Jr: it's used internally in some allocators, to be able to keep pages of memory with the program if no one else needs them, whether it's used in glibc's I don't know but that would be strange if not
 47 2014-04-29 05:58:20 <Luke-Jr> wumpus: I'm not sure what use that would be? C doesn't expose cache-like functionality O.o
 48 2014-04-29 05:58:56 <gmaxwell> use it for freed memory. "the OS can have it back if it wants it, otherwise I'll reuse this page.
 49 2014-04-29 05:58:59 <gmaxwell> "
 50 2014-04-29 05:59:48 <Luke-Jr> wouldn't reuse (removing the madvise) cost just as much as getting a new page? O.o
 51 2014-04-29 06:00:05 <wumpus> right; historically it has always been "when I claim this page from the OS, it stays with me forever" under UNIX (*yuck* SunOS), but Linux's software is somewhat smarter there
 52 2014-04-29 06:02:07 <wumpus> (but the other alternative "give the page back to the OS immediately" results in bad performance due to many syscalls, so that was avoided, madvise provides a compromise)
 53 2014-04-29 06:03:52 <wumpus> I'm not sure it's *that* useful for keeping caches, as you'd have to check every time before you access the page
 54 2014-04-29 06:06:42 <wumpus> that's some paranoid programming *does my buffer still exist? do I need to rebuild/reload it?*
 55 2014-04-29 06:08:38 <wumpus> it's more common in APIs where memory is limited and precious (such as GPU memory)
 56 2014-04-29 06:11:06 <wumpus> but I suppose these things are coming back, no one wants swap space anymore
 57 2014-04-29 06:11:31 <Luke-Jr> Linux doesn't work right without swap space :/
 58 2014-04-29 06:11:42 <Luke-Jr> it'll start OOM killing stuff even when there's free space
 59 2014-04-29 06:11:46 <Luke-Jr> free memory*
 60 2014-04-29 06:13:25 <venzen> i agree - have a laptop with 8GB RAM and without swap it would just go crazy sometimes - despite 2.5GB free mem
 61 2014-04-29 06:14:04 <Luke-Jr> heh, I have 16 GB RAM and I actually have to *use* swap :x
 62 2014-04-29 06:14:04 <venzen> and if you run out of mem - that is a crushing lockup :(
 63 2014-04-29 06:14:08 <wumpus> I tried encrypted swap but it was really, really slow and caused the system to hang for minutes at a time when swapping was needed, so now I'm running without swap and getting re-acquainted with the OOM killer
 64 2014-04-29 06:14:39 <Luke-Jr> wumpus: maybe a swap file on a ramdisk?
 65 2014-04-29 06:14:51 <gmaxwell> wumpus: hm! I've run w/ encrypted swap for years. Not noticed any issues like that. (also, I use tmpfs— so it's not like my swap doesn't get used).
 66 2014-04-29 06:14:53 <wumpus> Luke-Jr: but a ram disk is in ram... how would that help?
 67 2014-04-29 06:15:32 <gmaxwell> (unfortunately I just accidentally let my battery go dead in my laptop and restarted so I can't extract any useful numbers about how much swap I'm currently using)
 68 2014-04-29 06:15:42 <Luke-Jr> wumpus: it could convince Linux it has swap
 69 2014-04-29 06:15:50 <wumpus> gmaxwell: it worked fine *except* when actual swapping was needed, things became really really slow... and I don't care about slowless in general, but it's really that the mouse cursor hangs like it's 1994
 70 2014-04-29 06:16:38 <gmaxwell> yea, it might just be that I've basically never had it in that state.  Also— how were you doing the encryption?  Just a plain dmcrypt or did you have some kind of file on an encrypted fs?
 71 2014-04-29 06:16:45 <venzen> wumpus: sometimes linux just insists on using swap (low level or some app sys calls) and just having _some_ swap in a RAM disk is better than none at all in those cases
 72 2014-04-29 06:17:38 <Luke-Jr> once upon a time, I had Eligius running without swap. it'd get random stuff OOM killed all the time with several GBs RAM available
 73 2014-04-29 06:17:44 <wumpus> gmaxwell: I've also experimented with different kernel modules that impemet encryption (aes_x86_64 and such) but it didn't help either; but it still may be that more modern CPUs cope better as they have encryption instruction support
 74 2014-04-29 06:18:30 <wumpus> gmaxwell: using an entry in crypttabl (so that's dmcrypt, I suppose)
 75 2014-04-29 06:19:25 <wumpus> Luke-Jr: it very predictably kills my IRC client always :-)
 76 2014-04-29 06:19:34 <venzen> i just upgraded to lubuntu 14.04 and the system resource usage is a great improvement on vanailla 12.04, btw
 77 2014-04-29 06:19:43 <gmaxwell> wumpus: even without the encryption instructions it should still be much faster than the disk itself. (doesn't mean the kernel is handling it well though).
 78 2014-04-29 06:21:31 <wumpus> gmaxwell: agreed; i mostly wonder why the entire machine seems to wait for the swap operation, more than the slowness (I can live with a slower compile, but not if I can't do anything else in the meantime)
 79 2014-04-29 06:22:32 <Belxjander> wumpus: because the whole machine has MultiTasking waiting on the memory swap exception to complete
 80 2014-04-29 06:22:46 <Belxjander> wumpus: honestly no idea if that blocks SMP systems or not however
 81 2014-04-29 06:22:48 <Luke-Jr> wumpus: probably GTK's fault :P
 82 2014-04-29 06:22:48 <wumpus> and man, g++ sure uses a lot of memory while compiling... I probably should get a machine specifically for compling
 83 2014-04-29 06:23:17 <Luke-Jr> oh, my system slows to a crawl compiling LLVM..
 84 2014-04-29 06:23:35 <Luke-Jr> I think I had to give up on debug symbols for WebKit-based code too
 85 2014-04-29 06:23:43 <wumpus> Belxjander: hehe, swapping out the scheduler :)
 86 2014-04-29 06:24:12 <Belxjander> wumpus: I've actually HAD to change scheduler algorythms on a running multitasking system :)
 87 2014-04-29 06:24:56 <wumpus> Luke-Jr: whoa, does it use more memory if you want debug symbols? you'd say a compiler *always* needs those symbols
 88 2014-04-29 06:25:05 <Belxjander> 
ACTION was lucky as hell that it was all based on a shared memory system and the whole OS was based on a single pointer at a fixed location being readable for access to the kernel library... beyond that things would have gotten more complicated

 89 2014-04-29 06:25:57 <Luke-Jr> wumpus: yes
 90 2014-04-29 06:26:17 <Luke-Jr> wumpus: it uses a huge amount more, in fact - Gentoo advises NOT building WebKit/LLVM/etc with -g options
 91 2014-04-29 06:26:30 <_andares> jcorgan: what's the future of pull request #3652?
 92 2014-04-29 06:26:30 <Luke-Jr> wumpus: IIRC it's mostly the linker
 93 2014-04-29 06:26:35 <wumpus> Luke-Jr: but I like my debug data :(
 94 2014-04-29 06:26:45 <Luke-Jr> wumpus: me too, and I tolerate it for LLVM :p
 95 2014-04-29 06:26:56 <Luke-Jr> mainly because LLVM is responsible for BFGMiner crashing :/
 96 2014-04-29 06:27:54 <wumpus> Luke-Jr: ... so you're building a compiler from source? do you need the latest and greatest C++11 features, or some weird platform?
 97 2014-04-29 06:28:00 <Luke-Jr> (free software OpenCL is implemented with LLVM targetting the GPU…)
 98 2014-04-29 06:28:06 <Luke-Jr> wumpus: Gentoo builds everything from source
 99 2014-04-29 06:29:27 <wumpus> Luke-Jr: yes, Mesa uses LLVM to compile shaders for some newer GPUs, pretty cool, although writing a LLVM backend is much more work than the old TGSI framework, but GPUs have become thus advanced that they need a real compiler
100 2014-04-29 06:29:48 <wumpus> _andares: not going to be merged
101 2014-04-29 06:30:05 <wumpus> _andares: an address-based index doesn't belong at the level of bitcoind
102 2014-04-29 06:30:29 <wumpus> _andares: there have been ideas floating around about adding a UTXO query by address, but not the spent stuff
103 2014-04-29 06:31:02 <_andares> hm. I'm writing a program that needs to get a provably complete tx history for an address
104 2014-04-29 06:31:10 <wumpus> _andares: (comparatively spoken, a UTXO index by output would be pretty small, whereas indexing the whole block chain by address is crazy)
105 2014-04-29 06:31:23 <michagogo> cloud|9:28:31 <Luke-Jr> wumpus: Gentoo builds everything from source <-- Chicken-and-egg issue there, no?
106 2014-04-29 06:31:33 <_andares> wumpus: what's UTXO?
107 2014-04-29 06:32:04 <Luke-Jr> michagogo|cloud: the initial install has a tarball with just the compiler and package manager basically, which are rebuilt first
108 2014-04-29 06:32:08 <wumpus> too much extra code and it's only useful for block chain forensics, something that can be done exellently in a seperate tool (as a few do, like insight...)
109 2014-04-29 06:32:48 <wumpus> michagogo|cloud: the olde bootstrapping problem :)
110 2014-04-29 06:32:51 <michagogo> cloud|Luke-Jr: does "everything" include GNU/Linux?
111 2014-04-29 06:33:09 <Luke-Jr> michagogo|cloud: I'm not sure what you mean by that question.
112 2014-04-29 06:33:15 <michagogo> cloud|(I think I'm using the right name there)
113 2014-04-29 06:33:27 <wumpus> michagogo|cloud: yes, everything, every single tool, and the kernel as well
114 2014-04-29 06:33:28 <Luke-Jr> _andares: keep in mind that an address should only ever be used by 1 transaction
115 2014-04-29 06:33:44 <Luke-Jr> actually, Gentoo makes you build your own kernel. a bit annoying :/
116 2014-04-29 06:34:07 <_andares> Luke-Jr: I'm trying to build a reputation system, I have to make the anonymity tradeoff.
117 2014-04-29 06:34:12 <michagogo> cloud|So it's not just the compiler and package manager that the initial installer needs to give you
118 2014-04-29 06:34:12 <wumpus> michagogo|cloud: it's pretty elegant, and in a purist sense it's how open source is supposed to work, but I'm not patient enough anymore to run it
119 2014-04-29 06:34:50 <Luke-Jr> _andares: sounds like a bad idea. and it's security too, not just "anonymity" (which is actually "privacy", in the sense that making it a "tradeoff" is probably illegal under consumer protection laws..).
120 2014-04-29 06:35:02 <Luke-Jr> michagogo|cloud: hm?
121 2014-04-29 06:35:17 <_andares> Luke-Jr: fine, propose a better alternative or leave me alone.
122 2014-04-29 06:35:18 <Luke-Jr> wumpus: except for the memory-hungry builds, I just let it run in the background :p
123 2014-04-29 06:35:23 <michagogo> cloud|9:32:30 <Luke-Jr> michagogo|cloud: the initial install has a tarball with just the compiler and package manager basically, which are rebuilt first
124 2014-04-29 06:35:28 <Luke-Jr> _andares: PGP has one
125 2014-04-29 06:35:40 <michagogo> cloud|Sounds like it also needs to include the kernel and tools
126 2014-04-29 06:35:53 <_andares> Luke-Jr: I'm not aiming for web of trust, I want a seller feedback thing like amazon has for purchases.
127 2014-04-29 06:35:59 <Luke-Jr> michagogo|cloud: there's a livecd, but I usually just boot Kubuntu for that
128 2014-04-29 06:36:06 <Luke-Jr> _andares: that's a web of trust.
129 2014-04-29 06:36:33 <michagogo> cloud|Writing the first compiler can't have been fun
130 2014-04-29 06:36:46 <_andares> Luke-Jr: it is not, because I am only interested in cases where the transaction went badly.
131 2014-04-29 06:36:59 <michagogo> cloud|_andares: ...why?
132 2014-04-29 06:37:01 <jcorgan> wumpus: i understand #3652 isn't planned to be merged, but I'll keep it fresh with ongoing changes in current master if you ever change your mind :)
133 2014-04-29 06:37:02 <Luke-Jr> _andares: good luck getting a scammer to cooperate
134 2014-04-29 06:37:06 <_andares> anyway, this is getting off topic
135 2014-04-29 06:37:10 <michagogo> cloud|Positive feedback is also important
136 2014-04-29 06:37:16 <_andares> I'd be happy to continue the conversation but in a different channel
137 2014-04-29 06:37:17 <wumpus> jcorgan: yes, I don't mind; I wish bitcoind was more modular
138 2014-04-29 06:38:01 <Luke-Jr> michagogo|cloud: I'm still amazed that my father disassembled his entire compiler in order to merely customise its output format. and he can't handle Python..
139 2014-04-29 06:39:01 <SomeoneWeird> wat
140 2014-04-29 06:41:31 <Luke-Jr> (this was back when assembly was basically the only language)
141 2014-04-29 06:42:01 <arubi> there's a pun somewhere there ^^
142 2014-04-29 06:46:36 <michagogo> cloud|Luke-Jr: the question is, could he still do it today
143 2014-04-29 06:46:40 <wumpus> jcorgan: I proposed something once (https://github.com/bitcoin/bitcoin/issues/3440) to make it possible to have seperate modules that expose additional functionality, then the 'address based index' could just be a module that we pretend to not know about :)
144 2014-04-29 06:47:45 <wumpus> Luke-Jr: maybe you should show him the python bytecode :-)
145 2014-04-29 06:48:26 <jcorgan> i have read that PR in detail, and have much to say about it, but haven't been able to put my thoughts into a coherent post.  TL;DR--modularity good, message passing concurrency even better.
146 2014-04-29 06:48:44 <wumpus> 'look, this is almost like assembly too!'