< GitHub62>
[bitcoin] robegan21 opened pull request #7122: simple fix to mask command line arguments after they are copied (master...MaskCommandLineArgs) https://github.com/bitcoin/bitcoin/pull/7122
< GitHub170>
[bitcoin] gmaxwell opened pull request #7123: [WIP] Make trickle logic useful again, delay trickle when past upload limit. (master...actually_trickle) https://github.com/bitcoin/bitcoin/pull/7123
< phantomcircuit>
sipa, i guess i dont actually care that maxsigcachesize= isn't compatible since the chainstate isn't either
< sipa>
chainstate isn't?
< sipa>
between which versions?
< phantomcircuit>
sipa, 0.12.0 -> 0.11.2
< phantomcircuit>
because of obfuscation
< phantomcircuit>
oh actually it doesn't need to be changed
< phantomcircuit>
anybody who set it and goes to 0.12.0 will run out of memory
< sipa>
eh, we should change the name :)
< sipa>
grr, where did i put my glasses last noght?
< gmaxwell>
lol 100000 mb memory usage. darn non-typesafty of commandlines.
< sipa>
foubd!
< phantomcircuit>
gmaxwell, any opposition to adding a new checkpoint (keeping in mind that it's just to skip script validation now)
< gmaxwell>
Yes.
< gmaxwell>
As in yes, I'm opposed to doing that.
< gmaxwell>
The difference in validation time is basically negligible now; and to the extent that it's not we should address it by implementing the latching worth threshold shortcutting.
< phantomcircuit>
gmaxwell, failed to parse the last part of that sentence
< gmaxwell>
If the saved verification time isn't enough to justify implementing the real solution, then it's not enough to justify perpetuating the terrible screwup of checkpoints that causes material harm to the ecosystem in several ways.
< phantomcircuit>
ah right
< gmaxwell>
phantomcircuit: The signature shortcutting should just work based on a threshold difficulty; like skip verifying up to 2016 blocks from the tip on any chain whos total headers show work > X... but only on the initial sync (latching), don't skip in reorgs.
< gmaxwell>
So then we don't lose the optimization when we fail to constantly update some stupid constants, don't continue to give academics (for example) a reason to argue the consenus is actually hard coding blocks by us...
< gmaxwell>
And the only exposure it creates is a one time install time vulnerablity to attackers powerful enough that they could totally break everything anyways.
< phantomcircuit>
gmaxwell, my thinking on that was that the CScriptCheck objects could be queued on disk for processing later and then we could process them backwards
< phantomcircuit>
which also has the nice benefit of making it trivial to do the io bound stuff in parallel with the script checks
< sipa>
phantomcircuit: there's a much more elegant solution in just introduces an extra step in validation
< sipa>
and doing script validation in the background
< sipa>
but it's not trivial
< gmaxwell>
well you cannot verify scripts once the utxo are removed; unless you're extracting and storing the pubkeys.
< sipa>
they're stored in the undo files
< phantomcircuit>
sipa, i never thought about using the undo files for that
< sipa>
undo data may lack nLockTime though :s
< sipa>
so you can't validate bip65/bip112 that way :(
< gmaxwell>
sipa: does #7099 need rebase again?
< gmaxwell>
(I just rebased it before your comment)
< gmaxwell>
sipa: should I go update 7100 to remove the mruset?
< GitHub101>
[bitcoin] sipa closed pull request #6349: Constrain constant values to a single location in code (master...opt_defaults) https://github.com/bitcoin/bitcoin/pull/6349
< Luke-Jr>
MarcoFalke: eh, how did you do rpcport then?
< Luke-Jr>
… you didn't :/
< MarcoFalke>
Just left it how it was
< MarcoFalke>
yes
< Luke-Jr>
anything else missing?
< MarcoFalke>
no
< gmaxwell>
sipa: RE #7125 by 'No removal of the "1/4 blast to all" behaviour, as #7123 seems to offer a better replacement.' do you mean the sticky electricted trickling nodes? And darn wrt sorting the INV, I hadn't considered that the sort needs to be topological.
< sipa>
gmaxwell: i just mean "other pull requests exist that deal with this in a different way, let's not touch the code too much"
< GitHub26>
[bitcoin] luke-jr opened pull request #7127: Bugfix: Default -uiplatform is not actually the platform this build was compiled on (master...bugfix_uiplatform) https://github.com/bitcoin/bitcoin/pull/7127
< sipa>
without judging whether those others are a good idea, getting the non-controversial parts in first is priority
< sipa>
:)
< gmaxwell>
sipa: yea, I think that 7125 is easier to merge.
< gmaxwell>
I'd been debating to do per-peer timers in mine; it has some negative privacy effects when an observer has multiple connections to you (since they get the min of all of them); but it lets you have longer intervals with less effect.
< gmaxwell>
It also breaks up bandwidth usage more, which is probably good for usability.
< sipa>
gmaxwell: i would not be confortable with just per-peer timers if there was no bypass mechanism (like the 1/4 blast to all, or specific nodes that get everything instantly, ...)
< sipa>
as that would introduce inevitable delays on every hop
< gmaxwell>
In the per-peer, the 'delay per hop' is minimum (/second from minimum) of all those uniform numbers; e.g. ~N/2 fold smaller.
< sipa>
and the more independent paths (of the same length) there are between two nodes, the lower the delay would be
< GitHub170>
[bitcoin] luke-jr opened pull request #7128: Constrain rpcport default values to a single location in code (master...const_rpcport) https://github.com/bitcoin/bitcoin/pull/7128