< aj>
jtimon: re: getblockstats rpc tests, aren't tx sizes non-deterministic due to the signatures sometimes being a byte short due to a leading \x00 getting skipped due to DER encoding?
< jtimon>
aj: oh, right, that must be it...
< jtimon>
isn't there a way to solve that?
< sipa>
aj: how does thatmake them nondeterministic?
< sipa>
DER is deterministic
< sipa>
or you mean the low-S malleability?
< gmaxwell>
sipa: I think he thinks the signature is non-determinstic?
< aj>
sipa: running the same test case repeatedly doesn't generate the same R each time does it?
< gmaxwell>
it will if the transaction is the same.
< sipa>
aj: you can always choose to create a tx with a larger or smaller sig
< sipa>
by grinding
< sipa>
but if you have an actual tx, the serialization of that sig is fixed
< aj>
sipa: ack. but the test suite doesn't ensure it produces a deterministic chain, so txes will vary between test suite runs, yes/no?
< sipa>
no, our own signing is deterministic
< sipa>
you can't assume that other's signatures are - but the ones we create ourselves are
< aj>
sipa: okay, again, correct me if i'm wrong, but i'm seeing "self.nodes[0].getnewaddress()" giving different results between test_runner.py invocations, which (even despite deterministic signing) should result in non-deterministic tx sizes aiui
< sipa>
aj: oh, sure
< sipa>
if the the transaction isn't the same
< aj>
sipa: yeah. i wasn't following whether the implication was that there was meant to be deterministic generation of addresses/chains to have the test suite results be deterministic
< sipa>
aj: i just read some more context of the discussion, i wasn't aware you were talking about entire tests being deterministic
< bitcoin-git>
[bitcoin] promag opened pull request #11514: Iterate and remove nodes without container copy (master...201710-node-iterate-and-erase) https://github.com/bitcoin/bitcoin/pull/11514
< bitcoin-git>
[bitcoin] promag opened pull request #11515: Assert cs_main is held when retrieving node state (master...201710-node-state-guard) https://github.com/bitcoin/bitcoin/pull/11515
< bitcoin-git>
[bitcoin] practicalswift opened pull request #11516: crypto: Avoid creating of one-past-the-end pointer and performing a noop memset call (HMAC-SHA{256,512}) (master...one-past-the-end) https://github.com/bitcoin/bitcoin/pull/11516
< bitcoin-git>
bitcoin/master a44a215 Pedro Branco: Fix importmulti bug when importing an already imported key
< bitcoin-git>
bitcoin/master 808c84f MarcoFalke: Merge #11483: Fix importmulti bug when importing an already imported key...
< bitcoin-git>
[bitcoin] MarcoFalke closed pull request #11483: Fix importmulti bug when importing an already imported key (master...bugfix/fix-importmulti-bug) https://github.com/bitcoin/bitcoin/pull/11483
< cfields>
sipa: i forget your preference. You like a github review (batch), or individual comments?
< BlueMatt>
cfields: for the sake of (most peoples') inbox, batch
< cfields>
heh, fair enough. iirc someone complained about the batches because it screwed up their workflow.
< cfields>
hmm, maybe that was you, with your online/offline split?
< sdaftuar>
cfields: i'm thinking about adding a way to mock out addrman for the purposes of regtest-testing our outbound peer logic. does that sound like a reasonable thing to do?