< gmaxwell>
"I have made progress and while the current implementation is 20% slower than it could be it is still five times faster than before and think I will settle for that. Details are in my bitcoin-armv8 repo. It is reindexing my copy of the bitcoin blockchain at over double the pace."
< gmaxwell>
so we might want to steal that code. :P
< luke-jr>
theft is bad
< gmaxwell>
obviously I mean friendly stealing.
< luke-jr>
:P
< gmaxwell>
(I mean considering he's just transliterating our existing optimization to arm, I doubt he'd mind. :P )
< luke-jr>
eh? if we already have it, what is there to copy?
< sipa>
luke-jr: we don't have optimized code for ARM
< gmaxwell>
luke-jr: we have it for x86.
< luke-jr>
ah
< gmaxwell>
he's converting the x86 optimized functions to the arm analogs.
< gmaxwell>
sounds like he could use a bit of help in part of it too.
< gmaxwell>
(and also help with the build system parts)
< zib>
2.5 weeks to sync on the RPi instead of 5 :D
< gmaxwell>
zib: hey, that would be a pretty big improvement. :P
< zib>
Agreed. A bit more polished turd :D But I assume it's for all block validation not just reindexing?
< gmaxwell>
zib: yes, these speedups are for all block validation, reindexing is just a nice benchmark.
< zib>
Got it. Then it's pretty nice.
< gmaxwell>
For compact block relay the speedup should be pretty close to the function speedup in most cases, since its bottlenecked on hashing. (e.g. 5x)
< zib>
I need to figure out why the RPi can't handle encrypted swap (regardless of how heavy the encryption is)
< zib>
Want to have the swap encrypted on the full node + ln but it dies as soon as it start swapping.
< zib>
And you really need a swap on those little things :P
< gmaxwell>
zib: stick a fan on it.
< gmaxwell>
back in the day, some people would talk about 'windows quality hardware' -- basically windows crashed so much that you could release a pretty faulty system and people wouldn't notice since even if it crashed a couple times a week that was still 10x fewer than windows itself and not noticible...
< gmaxwell>
it seems the arm board vendors saw that and said "hold my beer". mobile devices end up heavily throttled due to being battery powered... I've never owned one that wouldn't crash or corrupt data under real load.
< gmaxwell>
(Novena comes closest to okay, but even that, if I put the libsecp256k1 tests on it in a loop, it eventually shuts off with a thermal overload, unless I add a fan)
< zib>
Didn't check if it's crashing from overheating but I guess it's possible.
< zib>
Better just run it inside the fridge.
< phantomcircuit>
zib, ensure you're power supply isn't browning out
< zib>
phantomcircuit: I thought that at first but tried 2 different official adapters
< phantomcircuit>
zib, ok but like monitor the 5v rail and see what happens at full load
< zib>
But like there's no problem if its a normal swap. I doubt the encryption brings it over the edge power wise.
< phantomcircuit>
also if you're swapping to an encrypted sdcard you're going to be sad
< sipa>
zib: raspberry pi in a fridge == device that consumes as much as a desktop PC with a few % of the computation power :p
< zib>
sipa: :D not a dedicated fridge
< phantomcircuit>
zib, doesn't matter a refrigerator is super inefficient
< phantomcircuit>
they're relying on the insulation
< zib>
I was just joking man :D
< gmaxwell>
hard to tell, poe's law for small computers.
< zib>
I like these tiny devices. Esp the RPi Zero. Got one for monitoring the wind+rain, one to turn off my sprinkler if there's been rain. One to open/close the front gate.
< zib>
And resin coated one to put on a bouey in the ocean. USB-connector rusted but the board was fine :D
< meshcollider>
i think you can make it happy just by changing the code
< luke-jr>
meshcollider: ?
< meshcollider>
remove the account alias
< luke-jr>
that'd reintroduce a bug
< meshcollider>
the account alias was removed for a reason
< meshcollider>
it should accept "*" but it should not accept account="*"
< luke-jr>
why not? that's exactly how you'd specify it..
< meshcollider>
as of 0.18 it would be specified with the dummy name or not at all
< luke-jr>
as of 0.18, it will hopefully be trusted_only
< meshcollider>
yes so just use trusted_only and not account
< meshcollider>
account -> dummy was already a breaking change so dummy -> trusted_only is fine
< luke-jr>
meshcollider: it's not supposed to be breaking existing code
< sipa>
accounts are gone, that already is a breaking change
< sipa>
one that was properly deprecated first
< luke-jr>
sipa: despite the name, it's not used for an account in this case
< sipa>
then there is no reason to use that name, right
< luke-jr>
sipa: the parameter was either null or "*" for "total wallet balance" calculated differently, or an account name
< luke-jr>
sipa: the reason to use the name is that existing callers use it
< sipa>
luke-jr: all callers that were using the account features had to change their code already
< meshcollider>
^
< sipa>
presumably that includes even this argument
< luke-jr>
but this doesn't use the account features, only the "account" parameter name
< luke-jr>
to indicate something that is not account-related
< meshcollider>
all uses of "account" parameter name have already been broken
< sipa>
because before we discovered it actually had an impact on confirmations, it was an account related thing
< meshcollider>
so dont add it back
< luke-jr>
sipa: it wasn't in this use case
< sipa>
this is purely incidental
< luke-jr>
hardly "discovered" when it was known/documented behaviour that people rely on
< luke-jr>
eg, I use it for my tax calculations
< meshcollider>
do you use account="*" ?
< luke-jr>
yes
< luke-jr>
oh, as opposed to positional parameters? not sure
< luke-jr>
looks like I am using positional, but that's not really the point
< meshcollider>
im not suggesting we break positional parameters
< meshcollider>
noone should be using any account named parameter now
< luke-jr>
if you were using named params, you HAD to use the name "account" for this
< meshcollider>
HAD to but not now
< meshcollider>
because accounts have been deprecated and removed the proper way, so people should migrate to the new name
< luke-jr>
there was no new name yet
< meshcollider>
it was called dummy
< meshcollider>
and you are changing to trusted_only
< luke-jr>
dummy never worked
< meshcollider>
either way, its not called accounts
< meshcollider>
yes so fix the behaviour of dummy, dont revert back to deprecated accounts :(
< meshcollider>
the name change dummy -> trusted_only is good though
< luke-jr>
so that there's literally no way for someone to write code that works with 0.16 and 0.17.1?
< luke-jr>
(using named args)
< meshcollider>
thats generally how breaking changes work, yes
< luke-jr>
there's a deprecation step to avoid that
< luke-jr>
and there's no reason to break in this case
< meshcollider>
keep the "dummy" alias as well as "trusted_only" then
< luke-jr>
why? "dummy" was never used for this
< meshcollider>
Oh I think I see what you mean. The account alias can be kept in the backport to 0.17.1 but not in master
< meshcollider>
dont add it to master
< luke-jr>
not sure if my ok got through before freenode d/c me, but no compat in master, just in 0.17 sounds reasonable
< meshcollider>
thanks :)
< bitcoin-git>
[bitcoin] luke-jr opened pull request #14781: [0.17] Bugfix: Correctly calculate balances when min_conf is used, and for getbalance("*") (master...bugfix_rpc_getbalance_acctstar-0.17) https://github.com/bitcoin/bitcoin/pull/14781
< bitcoin-git>
[bitcoin] luke-jr closed pull request #14781: [0.17] Bugfix: Correctly calculate balances when min_conf is used, and for getbalance("*") (master...bugfix_rpc_getbalance_acctstar-0.17) https://github.com/bitcoin/bitcoin/pull/14781
< bitcoin-git>
[bitcoin] luke-jr opened pull request #14782: [0.17] Bugfix: Correctly calculate balances when min_conf is used, and for getbalance("*") (0.17...bugfix_rpc_getbalance_acctstar-0.17) https://github.com/bitcoin/bitcoin/pull/14782
< bitcoin-git>
bitcoin/master 3615003 Luke Dashjr: net: Always default rpcbind to localhost, never "all interfaces"...
< bitcoin-git>
bitcoin/master d6a1287 Luke Dashjr: CNetAddr: Add IsBindAny method to check for INADDR_ANY
< bitcoin-git>
bitcoin/master 27c44ef Luke Dashjr: rpcbind: Warn about exposing RPC to untrusted networks
< bitcoin-git>
[bitcoin] laanwj closed pull request #14532: Never bind INADDR_ANY by default, and warn when doing so explicitly (master...rpcbind_explicit) https://github.com/bitcoin/bitcoin/pull/14532
< promag>
wumpus: #14756 is ready to merge, after jnewbery review the script interface remains the same
< bitcoin-git>
bitcoin/master d6cde00 João Barbosa: rpcauth: Improve by using argparse and getpass modules
< bitcoin-git>
bitcoin/master 708cbb1 Wladimir J. van der Laan: Merge #14756: Improve rpcauth.py by using argparse and getpass modules...
< bitcoin-git>
[bitcoin] laanwj closed pull request #14756: Improve rpcauth.py by using argparse and getpass modules (master...2018-11-improve-rpcauth.py) https://github.com/bitcoin/bitcoin/pull/14756
< promag>
ops, bug in master
< wumpus>
promag: ?
< bitcoin-git>
[bitcoin] promag opened pull request #14783: qt: Fix boost::signals2::no_slots_error in early calls to InitWarning (master...2018-11-fix-noslotserror) https://github.com/bitcoin/bitcoin/pull/14783
< promag>
wumpus: ^
< promag>
nothing serious
< promag>
maybe we should just call `noui_connect()` in bitcoin-qt?
< wumpus>
maybe? does it register any other things that might interfere
< wumpus>
I'm sure there was a reason, at some point, not to do that
< promag>
I'll leave as it is (as a minimal fix) but also leave that suggestion
< promag>
wumpus: I don't understand your last question
< promag>
what return value?
< wumpus>
so to be clear: the actual problem here is that the MessageBox slot gets called before the UI is set up far enough that any handlers are added?
< promag>
yes
< wumpus>
ThreadSafeMessageBox's return value
< wumpus>
it returns a bool
< wumpus>
depending on what the user chooses
< jonasschnelli>
I will miss the meeting again today (and probably next week).
< wumpus>
jonasschnelli: thanks for letting us know
< jonasschnelli>
getting up 3am is not very comfortable (currently in GMT+7)
< wumpus>
true!
< promag>
there are a 6 calls to InitWarning in init.cpp
< wumpus>
but non in that early stage
< promag>
at least 2
< promag>
in InitParameterInteraction()
< promag>
InitWarning(strprintf(_("Config setting for %s only applied on %s network when in [%s] section."), arg, network, network));
< promag>
InitWarning(strprintf(_("Section [%s] is not recognized."), section));
< promag>
these were added recently right?
< wumpus>
so, then, this was not a regression in #14708?
< wumpus>
probably... and never tested with the ui
< promag>
also guilty :(
< wumpus>
in any case, it's probably safest to convert those to LogPrintf and dn't mess with the slot handlers
< wumpus>
we're simply not able to show a message box or do anything else UI-wise at that point in the initialization process
< promag>
wumpus: yap
< promag>
wumpus: I'm checking older commits of that PR
< kallewoof>
jonasschnelli: welcome to my world lol
< wumpus>
maybe later on the UI code can be refactored so that that *is* possible, but that's a too large and risky change now
< promag>
looks like InitWarning was added in the end
< wumpus>
(e.g. there are some things concerning locale and language setup that only can run *after* configuration parsing, and I think using Qt things before that will currently cause bad things to happen)
< bitcoin-git>
[bitcoin] merland opened pull request #14787: Replace old instructions with links to official docs (master...issue14639) https://github.com/bitcoin/bitcoin/pull/14787
< bitcoin-git>
[bitcoin] ken2812221 opened pull request #14788: tests: Possible fix the permission error when the tests open the cookie file (master...fix-win-test-perm-deny) https://github.com/bitcoin/bitcoin/pull/14788
< bitcoin-git>
[bitcoin] MarcoFalke opened pull request #14789: qa: Run feature_block tests on p2p and rpc interface (master...Mf1811-qaBlockRpc) https://github.com/bitcoin/bitcoin/pull/14789
< bitcoin-git>
[bitcoin] kristapsk opened pull request #14790: Tests: Allow running rpc_bind.py --nonloopback test without IPv6 (master...testrpcbind) https://github.com/bitcoin/bitcoin/pull/14790