372300+ entries in 0.231s
punkman: also
the whole freenet wot dimension, you don't have
to connect
to assholes iirc
punkman: who knows what horrors lie in its source code
though
punkman: freenet design makes a lot more sense
than
tor
ascii_field: usg has been
trying
to phase out
the 'states'
thing since lincoln.
assbot: Logged on 27-11-2015 16:39:23; mircea_popescu:
this insanity is STILL with us, still "works" ?
jurov: yes, but was mapNextTx deallocated at
that
time?
ascii_field: iirc (i don't have
the box here) i invoked RemoveFromMemoryPool
jurov: did you? i'm still putting it all
together
jurov: the jettison patch had it
the other way
ascii_field: but is sadly lying around somewhere deep in
the ml archive
jurov: CInPoint refers
to CTransaction,
thus mapNextTx should be erased before mapTransactions
ascii_field: i forget
the name, it is not in
the vpatch megacanon
ascii_field: jurov, phf: i recommend getting
that patch of mine which lets you request EXACT bytes of heap used
total at ANY
TIME
ascii_field: but if other folks want
to
take
their
turn in
the manhole - sure, why not...
ascii_field: my original purpose for asking nubbins for
The Book, if anyone remembers, was so
that i could rewrite.
ascii_field: the whole
thing is an epic
turd and my desire
to ever work with it again borders on zero
phf: ah and
that does
the crash?
phf: there's
this CTransaction::RemoveFromMemoryPool, anybody
tried calling
that in a loop?
ascii_field: but if you do invoke
the destructors (as i later did) - you crash.
jurov: or why does it not destroys
the
transactions
ascii_field: phf:
the goal of 'zapmempool' was
to prove
that mempool accounts for all of
the bloat other
than what is accounted for by
the 300bytes/block eternal index (another idiocy which will prevent pogoization unless a massive mechanism is rewritten beyond recognition)
jurov: and how does
the map know
there are outstanding references?
jurov: phf yes but
that were stack operations. i
then filtered
them
ascii_field: i
tend
to associate
them in my head, both are of
the heathens
ascii_field: and yes, jurov was right re: it being a std:: crapolade, rather
than boost
ascii_field: now if you feel like it, reproduce
this other
thing i did
jurov: yes just reproduced it -
the destructors are not called when map is cleared
ascii_field: because
the data structure appears
to be an interdependent ball of shit
ascii_field: in sept. i did an experiment (can't recall if i bothered
to explain it here) where i forced
the cblock crapola
to deallocate, in
the correct way (without using an iterator)
ascii_field: every
tx received eats a measurable qty of ram which is NEVER reclaimed
phf: jurov:
the goal is
to get bitcoind working with pogo's limited memory.
the problem is
that a running bitcoind grows in memory use as a result of normal operations. we know
that some processes claim a lot of memory by design, like mempool, so first step is
to get a reliable way of cleaning out mempool. ascii wrote
that patch, but discovered
that in practice zapmempool doesn't reduce memory use.
assbot: Logged on 19-08-2015 14:32:04; asciilifeform:
thestringpuller:
this patch does not work as described, on account of boost idiocy (removal of items from
the hash does not invoke
their destructures! believe)
phf: what you mean by "narrowed
the problem so"? you mean boost, or "zapmempool
to baseline", or?
jurov: you see. if i knew why alf narrowed
the problem so, would help me much
phf: jurov: by
the way, in addition
to static
things,
there's also
top level globals
jurov: O.o how is it possible
to have static variable inside if statement?
phf: jurov: can paste file:line for
those?
jurov: just found
there are several static maps and sets
tucked in misc functions,
too...
☟︎ phf: last conversation on
the subject was mp saying basically, people are welcome
to investigate and plug
that hole
phf: but for all i know
there's all kinds of other sources of leak.
the general idea is
that btcd without mempool memory use should stay constant when
the blockchain is not growing. any kind of memory growth is from mempool adding more
transactions. so it stands
to reason
that a fully working zapmempool should always bring btcd back
to some baseline number, but
that's not happening.
phf: jurov: clear not working is my broken
telephone of
that ascii was saying in
the logs. i've only gotten as far as making sure
that
the code ~seems~
to be working, hence my conclusion, which is
that mempool is somehow not just mapTransactions & mapNextTx
jurov: ikr? but i'd like
to see some closer dissection. like, how did you determine clear() does not work
phf: jurov:
the overal goal
though is
to flush
the mempool, but simply measuring
the memory between zapmempool shows
that
the patch specifically doesn't do it.
there's either additional source of leak, or
there's a leak in mempool, or, and
that's
the most likely case,
the patch in question doesn't
touch al lthe places where mempool has data
phf: jurov: you know how it is here,
there's one small problem, gets elevated
to shitgnomes in all
the
things!!1 status
☟︎ jurov: are you
telling me boost modifiest std:: stuff behavior?
phf: jurov: so original interpretation was
that mapTransactions.clear(); doesn't deallocate individual instances (i.e. boost doesn't do
the right
thing), attempting
to iterate over mapTransactions and do explicit del on instances results in crash. so
the new assumption was
that mapTransactions instances are somehow used elsewhere. i don't
think
there's much else insight on
the subject
trinque: damn christians wont get uppity about anything other
than
than abortions.
BingoBoingo: Well, not confirmed courpses yet. Local news has stopped referring
to
the officers
though as "wounded" and has
transitioned
to "shot"
to describe
their condition
BingoBoingo: jurov: deedbot- didn't get
the
title update.
assbot: Logged on 19-08-2015 14:32:04; asciilifeform:
thestringpuller:
this patch does not work as described, on account of boost idiocy (removal of items from
the hash does not invoke
their destructures! believe)
jurov: mod6, asciilifeform,
thestringpuller: can you please enlighten me how does boost affect mapTransactions and its clear() function? i'm reading
the code and don't see it
jurov: felipelalli: worst what can happen is.... someone will
tell you something
mircea_popescu: this is
teh republic, if you don't do it it ain't getting done. we've neither a legion of mindless drones
to "do" stuff badly nor a printing press
to devalue everyone's ownership
to "pay" for
that "work".
felipelalli: mircea_popescu,
Thank you for encouraging! But I feel I have read / study more
the logs before feel comfortable
to write
to Qntra. Qntra.net is like my oracle!
trinque: speaking of which, why
the hell is deedbot gone again
mircea_popescu: i was kinda alone for half hour yest while assbot and everyhone deserted
the chan
jurov: i came and noone was
there on any chan
BingoBoingo: Just like 2012, china was a
thing beyond Asicminer derping about shares
BingoBoingo now REALLY feeling like
this is fall 2012 all over again
adlai: volume has gone negative in china; vwap ceases
to be a
thing
gribble: Bitfinex BTCUSD
ticker | Best bid: 360.01, Best ask: 360.17, Bid-ask spread: 0.16000, Last
trade: 360.2, 24 hour volume: 44791.1190569, 24 hour low: 343.02, 24 hour high: 367.0, 24 hour vwap: None
mircea_popescu: the "commenting community" is basically a fabrication. he'll "engage" it when it's
the usg's own indians doing it and
then only.
BingoBoingo: From comments on linked page: "It is certainly
telling
that despite close
to 200 comments on
the
two articles on
the MTStandard
that address
this change, Mr. McCumber has not engaged with
the commenting community, nor has he
taken questions from
those posted and addressed
them."
mircea_popescu: in other news, "files included in a list of known child porn file hashes from
the police database", seriously ?
punkman: @freenodestaff We are again experiencing connectivity problems
to some servers due
to DDoS attacks. Please bear with us while we ride it out.
mircea_popescu: "The continued accretion of fiscal crises across
these governments should be expected
to continue
to further weaken
their already anemic vestiges of autonomy." <<
this needs
to be 5 words.
ben_vulpes: "the continued accretion of fiscal crises across
these governments should be expected
to continue..."
pete_dushenski: my delving into french literature
thus far has been pretty limited, a bit of verne and exupery mostly.
brg444: I'm guessing you also dared
take
the plunge into
the literature as well? (Bastiat, etc.)
brg444: that's great. I guess
that's better
than 99% of english speaking canada can pretend
to
pete_dushenski: and now, as i speak it out loud
to mein heir on a daily basis, i'm practicing more
than i have since i was 10 years old !
pete_dushenski: i also have a couple friends who also attended french immersion, but
through
to grade 12, and we occasionally exchange a few sentences, which is better
than nothing :)
pete_dushenski: that, plus some spanish classes in high school helped
to flesh out my latin-language base.