< jonatack>
thanks again, i hadn't realized that the ci cases (the /$CASE part) were in qa-assets and i hadn't git pulled qa-assets to update it, this did it: FUZZ=netaddr_deserialize seed=3814760663 src/test/fuzz/fuzz ../qa-assets/fuzz_seed_corpus/netaddr_deserialize/1bc91feec9fc00b107d97dc225a9f2cdaa078eb6 ... cheers aj
< jonatack>
er, this: FUZZ=netaddr_deserialize src/test/fuzz/fuzz ../qa-assets/fuzz_seed_corpus/netaddr_deserialize/1bc91feec9fc00b107d97dc225a9f2cdaa078eb6
< jonatack>
MarcoFalke: fwiw, this is a case where the ci fuzzer found an issue that i did not see locally, even after an hour of fuzzing on a laptop
< jonatack>
(without updating qa-assets)
< bitcoin-git>
[bitcoin] jonatack opened pull request #22056: doc: describe in fuzzing.md how to reproduce a CI crash (master...fuzzing-doc-describe-how-to-reproduce-ci-crash) https://github.com/bitcoin/bitcoin/pull/22056
< MarcoFalke>
jonatack: Link, please
< MarcoFalke>
Make sure to compile with all sanitizers, if they are needed
< jonatack>
yes compiled with everything needed and reproduced instantly once i updated qa-assets
< jonatack>
(duh i guess, but good to refresh memory)
< MarcoFalke>
ok that one is a logic bug and doens't need sanitizers
< MarcoFalke>
What command did you use to do "an hour of fuzzing on a laptop"
< MarcoFalke>
?
< MarcoFalke>
With sanitizers enabled fuzzing is slower by a factor of ~10, so it could be that you only got 6 minutes of "real" fuzzing
< jonatack>
--with-sanitizers=fuzzer,address,undefined,integer and then the dumb version without the corpora: FUZZ=netaddr_deserialize src/test/fuzz/fuzz
< MarcoFalke>
Interesting. Let me try the same without sanitizers
< wumpus>
anyone running testnet nodes: can you please pastebin and send me the output of "bitcoin-cli getnodeaddresses 0 | grep -E '[a-z0-9]{56}\.onion'" , i'd like to make a PR that replaces the seed nodes for testnet, just going to pick a few
< jonatack>
(tested as connecting and staying connected at the time)
< wumpus>
jonatack: thanks! from last week is great, i'll try them again and if they're still up they're at least reasonably stable :)
< jonatack>
wumpus: good idea! yes, they were the pick of the moment out of ~70 testnet peers, then after that there were ~250 :p
< promag>
I'm getting "error: option 'cf-protection=return' cannot be specified on this target" when building boost depends on macos arm64 - anyone can reproduce?
< wumpus>
12 out of 13 of the ones in the commit itself are still connectable
< MarcoFalke>
jonatack: I couldn't run into a crash with 1kk iterations in a single thread
< MarcoFalke>
With "-use_value_profile=1 -workers=9 -jobs=9" and a folder to share the corpus, it took 400k iterations
< MarcoFalke>
Same as above without use_value_profile, it took 1.1kk iterations for the first crash
< MarcoFalke>
Same as above (use_value_profile=1) witout a shared corpus, it took 1.5kk iterations for the first crash
< MarcoFalke>
Result for this experiment (N=1): Use a folder to share the corpus, Use more CPU, -use_value_profile=1
< bitcoin-git>
[bitcoin] laanwj opened pull request #22060: contrib: add torv3 seed nodes for testnet, drop v2 ones (master...torv3-testnet-hardcoded-seeds) https://github.com/bitcoin/bitcoin/pull/22060
< wumpus>
macosx CI failing: "Warning: Calling `cellar` in a bottle block is deprecated! Use `brew style --fix` on the formula to update the style or use `sha256` with a `cellar:` argument instead."
< bitcoin-git>
[bitcoin] hebasto opened pull request #22063: build: Use Qt archive of the same version as the compiled binaries (master...210525-version) https://github.com/bitcoin/bitcoin/pull/22063
< bitcoin-git>
[bitcoin] practicalswift opened pull request #22065: Mark `CheckTxInputs` `[[nodiscard]]`. Avoid UUM in fuzzing harness `coins_view`. (master...nodiscard-CheckTxInputs) https://github.com/bitcoin/bitcoin/pull/22065