< bitcoin-git>
bitcoin/master b3c4d9b Sebastian Falbesoner: test: rename test suite name "tx_validationcache_tests" to match filename
< bitcoin-git>
bitcoin/master 2755b2b fanquake: Merge #18010: test: rename test suite name "tx_validationcache_tests" to m...
< bitcoin-git>
[bitcoin] fanquake merged pull request #18010: test: rename test suite name "tx_validationcache_tests" to match filename (master...20200127-test-adapt-test-suite-names-to-file-names) https://github.com/bitcoin/bitcoin/pull/18010
< provoostenator>
valgrind
< provoostenator>
Update on the valgrind mystery in #15382; it appears it doesn't like "echo"
< bitcoin-git>
[bitcoin] Bushstar opened pull request #18012: GBT segwit rule in RPC error msg missing single quotes (master...patch-5) https://github.com/bitcoin/bitcoin/pull/18012
< provoostenator>
echo also handles quotes differently between Windows and Linux, so I probably need an alternative incantation for that test anyway.
< fanquake>
Qt is ditching LTS releases for non-commercial users, and it looks like a "Qt Account" will soon be required to even download the open source packages: https://www.qt.io/blog/qt-offering-changes-2020
< fanquake>
Ooh actually, should still be able to download the source with no account. Looks like it'll just be for the installers?
< luke-jr>
fanquake: hmm, LTS only for commercial users sounds like pretty reasonable funding for open source
< luke-jr>
they should probably think of a more reasonable price for end users tho (not sure how that will mesh with open source licensing)
< luke-jr>
I guess there are practical complications of such a policy for a library ^^;
< luke-jr>
is there a reason #16507 wasn't backported? cc fanquake instagibbs
< achow101>
and maybe #17261, tho meshcollider needs to ack it
< gribble>
https://github.com/bitcoin/bitcoin/issues/17261 | Make ScriptPubKeyMan an actual interface and the wallet to have multiple by achow101 . Pull Request #17261 . bitcoin/bitcoin . GitHub
< fjahr>
luke-jr: maybe wrong pr? That one is already merged.
< luke-jr>
fjahr: not to backports
< fjahr>
luke-jr: ah ok, i thought you meant rtm :)
< luke-jr>
rtm?
< jeremyrubin>
ready to merge
< gwillen>
achow101: (or anybody) -- is it possible to have a PSBT with all inputs signed and yet for some reason we can't finalize and extract it
< gwillen>
(and is that "in theory" or in practice?)
< achow101>
no in theory, yes in practice :p
< achow101>
In theory, the finalizer should be pretty dumb
< gwillen>
what are the kinds of cases where this can happen in practice?
< achow101>
but in practice, it verifies the transaction before spitting it out, so in practice, if you have an invalid signature, it won't finalize
< gwillen>
ahhh, interesting
< gwillen>
is that the only case you're aware of?
< achow101>
yes
< achow101>
other cases are usually that you've done something subtly wrong and it isn't actually signed
< sipa>
or the scripts used are not understood by the finalizer
< sipa>
so despite having enough signatures present, the finalizer doesn't know how to combine them into a full satisfying witness
< gwillen>
I guess actually, from my perspective if I see that an input is "signed" that already means the signature parts have been combined
< gwillen>
and as far as I can tell if all inputs are "signed" in that way, then finalization looks like it can't fail
< gwillen>
nor extraction
< gwillen>
since PSBTInputSigned declares true if it sees a final_script_sig or a final_script_witness
< gwillen>
and extraction does nothing but copy those fields out, it does not check anything as long as they exist (although presumbly broadcast will fail if they're bad)
< sipa>
gwillen: finalization is creating the final_script_sig etc; if those already exist, then extraction will never fail
< sipa>
extraction in theory could run a script verifier first; there should be enough information in psbt to do that
< achow101>
sipa: is there a command line descriptor to script compiler somewhere?