24900+ entries in 0.053s
mod6: <+ascii_modem> bdb begs for the axe << yes. it's awful.
mod6: if you still don't have any spare time, we'll figure out something.
mod6: no sweat. we're not quite at that point anyway. as I was eluding to in the Address & ToDo, I'll update #b-a at some point during the month after we get this other leak sorted.
mod6: we appreciate any and all help on this endeavor
mod6: feel free to hit either ben_vulpes or me anytime if you wanna collaborate on that.
mod6: let him know that he's about to send a 40 BTC tx fee. there are improvments that can be made perhaps
mod6: so, well, i dunno, i guess "works as designed" or something, but a least do the guy a favor of a reach-around
mod6: yeah, but i just read too that if you don't set your change outputs, you'll end up sending like a huge tx fee.
mod6: of course, as always, we welcome any help that anyone wants to give to achieve these goals.
mod6: i'll stay the course on v0.5.3 and what we set out to do in january. will revisit this later in the month as was the plan anyway.
mod6: fair enough. perhaps this is not quite the urgent matter that I thought it might have been.
mod6: it may not be. i had a feeling that it is widely used for some reason.
mod6: i might just be dumb though. eheheh.
mod6: probably not. i was just hoping to grab that stuff, and get it back-ported quickly.
mod6: if nothing else, might just be a lot of work. perhaps do-able, just not a quick/easy way.
mod6: spent, probably a total of 16 hours on it. didn't get very far. and seeing as though the raw transactions api is 4 versions beyond (or something) v0.6.0, it might be very hard to do it that way.
mod6: when trying to back port something i thought would be as simple as import/dump private keys, it turned out that the changes were indeed far reaching in the code.
mod6: well, that maybe have been a poor choice of words.
mod6: I'm not sure there is any easy way to back-port the raw tx api
mod6: currently, the plan is to attempt to fix the leak that causes the R.I. to crash during full block sync. but we might end up rolling forward to version .7 and having to re-implement the 7 patches we currently have.
mod6: ok. just wanted to make sure I understood :)
mod6: (bullet point 4): fix behavior where blocks are downloaded from a single peer (what is the desired behavior here? - should be pretty much ripped off from SotA torrent trackers.)
mod6: ok, i haven't looked into the part yet where it has been somehow spotted that the blocks seems to get pulled from one peer, instead of many. if p2p seeds refers to a way to get blocks in more of a torrent fashion, i agree.
mod6: <+ascii_modem> neh he snipped whole thing << yeah.
mod6: so now it keeps the BDB TX logs down to < 1Mb each, and I've never seen more than 60 of them in the ~/.bitcoin/database/ dir before being flushed out
mod6: then after working on different things for a while, and looking at what is currently set in HEAD, went on to trim that shit out
mod6: at first, I just set max locks to 40000, and nothing else, but ended up with liek 265Gb of BDB Tx logs
mod6: and DB_LOG_AUTO_REMOVE
mod6: set DB_TXN_WRITE_NOSYNC
mod6: set set_lg_max to 1M
mod6: changed dbenv.set_lk_max_locks to 40000
mod6: changed dbenv.set_lk_max_objects to 40000
mod6: hopefully, won't be too bad.
mod6: foundation got a lot accomplished in December, but still have at least one defect to eradicate.
mod6: it goes! been busy. :)
mod6: cazalla: no worries. let me know if you guys have any questions.
mod6: <+kakobrekla> smoke and a pancake? << haha
mod6: oh, good question cazalla. i don't know. might have been.
mod6: haha Jason saw a landing strip
mod6: someone kicked their rig back on
mod6: asciilifeform: are you going to b-a?
mod6: im waitin on Bitcoin Motivation 101
mod6: I still have a bit more testing to complete before the end of the month, but hopefully will get some additionally time over the remaining 13 days to debug more.
mod6: mike_c: yeah, should have said "stem from or usage of BOOST" :]
mod6: obv. a lot smaller and only seems to grow up to ~60 log files until it flushes them out.
mod6: here's a snippit of what my .bitcoin/database dir looks like with the BDB patch compiled in:
mod6: i'll find out, one way or another.
mod6: well, indeed, that could be the case.
mod6: And perhaps could squash that bug at the same time. I need to do more work with gdb while it's sync'ing up to try to get a more accurate picture.
mod6: I just have a gut-feeling that perhaps as discussed in here before some of these leak problems stem from BOOST, which will be a pita to remove, but a requirement.
mod6: But yeah, it's still has some leaks in there somewhere, as the R.I. died at least 2 different times from 1->current block, but no more than 3 times. Eitherway, this needs to get resolved. But at least we're headed in the right direction.
☟︎ mod6: here's a snippit of what my 300Gb disk looked like with only the dbenv.set_lk_max_locks looked like:
mod6: So I looked into what they have set up in v0.9 and it more closely resembles that second dpaste above. If you guys would like to review a bit before I post these to the list, that'd be awesome.
mod6: I was able to get upto the current block 2 different times with patches 1->5 (which includes ascii's patches, ben's patch, and my removal of checkpoints) with two seperate bdb configs. The first one just changed this: dbenv.set_lk_max_locks(40000); But ended up with a huge amount of BDB tx log files.
mod6: This month I've been doing testing with the patched version of the reference implementation. Main object here is to get from genesis block to current block, while being able to get past the wedge block 252450. I was able to do that, but not only did we test with ascii's patches and ben's UPNP patch, I added a patch (
http://dpaste.com/1G3XY64) that removes the checkpoints, and a patch (
http://dpaste.com/1K5M2TN) to configure BDB so the R.I. won't
mod6: root@ip-172-31-10-219:/mnt/btc-dev/sandbox/bitcoin-bitcoin-a8def6b/src# ./bitcoind -rpccommand getblockcount
mod6: bsod @ heathrow ; hilarity ensues
mod6: i imagine these guys just sit there allday looking at pornhub and playing flappy bird on their new galaxy 5s IF they even show up to the session
mod6: those ten probably just hit the wrong voting button. margin of error or something.
mod6: return the favor: leave them a 2lb. of tripe and an xmas card on their doorstep
mod6: <+ben_vulpes> decimation: disk storage is the *bane* of bitcoin testing. << for sure.