< bitcoin-git>
[bitcoin] sipa opened pull request #18551: Do not clear validationinterface entries being executed (master...202004_fix_validation_notify_clear) https://github.com/bitcoin/bitcoin/pull/18551
< fanquake>
Have moved #15774 from 0.20.0 to the future milestone. Shouldn't be anything more than needs doing there for this release.
< fanquake>
Moved #17020 from 0.20.0 to 0.21.0. While those might all be nice improvements, nothing specific that should be blocking 0.20.0 at this stage.
< aj>
fanquake: the configure test doesn't check that particular call, i guess? but _mm_set_epi64x is supposed to be part of sse2 so seems pretty weird
< sipa>
the configure check is irrelevant
< fanquake>
sipa: btw can you block SatoshiNakamotogmx
< sipa>
that's testing whether your compiler supports that function, which it clearly does (otherwise the code would not have compiled)
< sipa>
the problem is that it's calling _mm_set_epi64x during global initialization, before any runtime checks are done
< sipa>
that's a bug, i think, because it would definitely fail on a system without sse2
< sipa>
(i still don't understand why it fails on *his* system, because it supports sse2...)
< aj>
sipa: the backtrace includes sha256_shani::Transform, which shouldn't get referenced unless have_sse4 is set unless i'm misreading?
< sipa>
fanquake: done
< sipa>
aj: i think it may just be "global initializer for"
< fanquake>
sipa: thanks. I couldn’t quite make out the Adobe photo..
< aj>
sipa: ah, yeah, it also references line 19 which is INIT0 def
< sipa>
aj: trying something
< bitcoin-git>
[bitcoin] sipa opened pull request #18553: Avoid non-trivial global constants in SHA-NI code (master...202004_no_global_init_shani) https://github.com/bitcoin/bitcoin/pull/18553
< sipa>
ah, what i suspect is going on: even though the intrinsic is an SSE2 one, it is executed in a compilation unit compiled for SSE4+SHA-NI, which means it's not impossible that that intrinsic is actually compiled to an SSE4 instruction for him
< sipa>
and then executed at startup before checking that SSE4 is available
< fanquake>
sipa: would that be as part of some random usage before AppInitSanityChecks() is called?
< sipa>
fanquake: no, a global initializer
< sipa>
before main() runs
< fanquake>
right, I thought one of those required some randomness. Will have to look again.
< sipa>
fanquake: no, look at the code in sha256_shani.cpp
< sipa>
the initializer for MASK, INIT0, INIT1
< sipa>
they require running code
< sipa>
there are global initializers for randomness too which are annoying, but that's unrelated to this
< fanquake>
ah ok, that's what I was thinking off. I see the calls in sha256_shani
< aj>
sipa: hmm, that makes sense, though not seeing what sse4 or avx instruction it could be. there's some "more efficient" variants in sse3 that could do it maybe
< bitcoin-git>
bitcoin/master d7092c3 Luke Dashjr: QA: Test that change doesn't turn into non-change when spent in an avoid-r...
< bitcoin-git>
bitcoin/master 2952c46 Luke Dashjr: Wallet: Replace CAddressBookData.name with GetLabel() method
< bitcoin-git>
bitcoin/master 7a2ecf1 Luke Dashjr: Wallet: Change IsMine check in CWallet::DelAddressBook from assert to fail...
< bitcoin-git>
[bitcoin] MarcoFalke merged pull request #18546: Bugfix: Wallet: Safely deal with change in the address book [part 2] (master...bugfix_addressbook_change) https://github.com/bitcoin/bitcoin/pull/18546
< bitcoin-git>
[bitcoin] mrwhythat opened pull request #18554: wallet: ensure wallet files are not reused across chains (master...wallet-file-reuse-prevention) https://github.com/bitcoin/bitcoin/pull/18554
< bitcoin-git>
[bitcoin] MarcoFalke opened pull request #18555: wallet: Abort when assumption fails in debug mode (master...2004-walletAssumeAbortDebug) https://github.com/bitcoin/bitcoin/pull/18555