< sipa>
if i try to print the values involved it goes away...
< sipa>
"test/functional/wallet_hd.py --valgrind" e.g.
< jonatack>
sipa: the tests run cleanly for me with --valgrind
< jonatack>
(clang 9 debian)
< sipa>
i have it both on ubuntu 21.04 and 20.10
< sipa>
gcc 10.2.0-13ubuntu1 and gcc 10.2.1-20ubuntu1
< jonatack>
tried with wallet_basic, wallet_hd, and wallet_multiwallet, using the --valgrind test runner option and also `valgrind test/functional/wallet_...`
< jonatack>
i'll try with gcc 10.2.1
< bitcoin-git>
[bitcoin] theStack opened pull request #21366: refactor: replace util::Ref with std::any (C++17) (master...202012-replace-util_ref-by-std_any) https://github.com/bitcoin/bitcoin/pull/21366
< sipa>
trying to see if i find something with ubsan
< * jonatack>
sees the good ol' txmempool.cpp:897 bogus warning scroll by...
< sipa>
nothing with ubsan
< jonatack>
sipa: ok, saw a bunch of FuzzedSock warnings as achow101 had mentioned, and then test/functional/wallet_hd.py --valgrind reproduces your issue
< jonatack>
gcc (Debian 10.2.1-6)
< jonatack>
Conditional jump or move depends on uninitialised value(s), CWallet::ScanForWalletTransactions(uint256 const&, int, std::optional<int>, WalletRescanReserver const&, bool) (wallet.cpp:1822)
< sipa>
yea
< sipa>
and it disappears whatever i do trying to investigate it
< bitcoin-git>
[bitcoin] MarcoFalke opened pull request #21370: Use C++11 member initializer in CNodeState (master...2103-netCNodeStateRefactor) https://github.com/bitcoin/bitcoin/pull/21370
< vasild>
about fuzzing - if the source code or the test is changed, does that make the fuzz seed corpus just a pile of random bytes, as useful as any other pile of random bytes?
< vasild>
my understanding is "yes"
< jonatack>
jonasschnelli: reproduced locally with test/functional/feature_nulldummy.py --valgrind
< bitcoin-git>
[bitcoin] jonatack opened pull request #21373: test: extend rpc timeout in feature_nulldummy to fix CI timeouts (master...fix_feature_nulldummy_test_timeout) https://github.com/bitcoin/bitcoin/pull/21373
< dongcarl>
Question for people who understand license better than I do: say I want to copy a few lines verbatim from an MIT-licensed repository, do I have to attribute? How?
< real_or_random>
dongcarl: is your project MIT-licensed too?
< dongcarl>
real_or_random: Yup
< real_or_random>
I'd just add the person to the copyright notice of the source file
< real_or_random>
you could also be more specific and say "these lines are written by" or something.
< dongcarl>
real_or_random: I see... What if I modified it a little? :-)
< real_or_random>
"/* based on ... written by ... */"
< dongcarl>
Cool cool, thanks real_or_random!
< real_or_random>
it's not clear if you need to do this in this detail but it probably does not hurt and it's nice towards the authors
< dongcarl>
True!
< real_or_random>
and now that I think about it, it's probably better than adding the name to the copyright header because that may be misunderstood
< real_or_random>
it looks like the person contributed to that file, which is not entirely true
< dongcarl>
Right, I'll be sure to be more explicit
< real_or_random>
if you want to be entirely correct, copy also the "copyright" line from their license file then.. because this is what the license requires
< real_or_random>
"The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software." (where the permission notice is already there in your repo)
< real_or_random>
a "few lines" are probably not substantial but who knows ^^
< dongcarl>
Ah, the project I'm copying from only _is_ a few lines, so most likely it is substantial. Good to know!
< real_or_random>
ok yeah. ^^ I mean, in that case it may be simpler to have it in a separate file but that shouldn't make a difference in the end
< real_or_random>
(this answer ended up more complex than I wanted it to be ^^)