< aj>
fjahr: your bench results on #16902 seem very strange. are you sure you're benching the right commits? for a in e6e622e5 89fb241c; do git checkout $a && make -j5 >/dev/null 2>&1 && echo "Testing $a" && bench/bench_bitcoin -filter=VerifyNestedIfScript -evals=200; done # to be sure maybe?
< fjahr>
aj: I know, I had checked it multiple times already because of that. First cleaned up with `make distclean && ./autogen.sh && ./configure` and then ran `for a in e6e622e5 89fb241c; do git checkout $a && make >/dev/null 2>&1 && echo "Testing $a" && src/bench/bench_bitcoin -filter=VerifyNestedIfScript -evals=200; done`
< sipa>
any debug/optimization flags that may matter? sanitizers?
< sipa>
ah, no
< aj>
fjahr: no VMs? what cpu?
< fjahr>
no vm, Intel(R) Core(TM) i5-6287U CPU @ 3.10GHz
< sipa>
well, it's still faster
< sipa>
that's really all we care about
< luke-jr>
was it SHA512 that caused the entire CPU to slow down if we used the hardware optimisation? I wonder how we'd handle if a future CPU solved that..
< luke-jr>
I guess we could benchmark at startup
< sipa>
luke-jr: AVX512 on supported systems; we don't use those instructions
< sipa>
(also hardly.anything supports them)
< luke-jr>
sipa: I was just pondering how/if we should, when/if Intel/AMD make a CPU that can use them effectively
< luke-jr>
having a "slow down the CPU" implementation on current CPUs seems to sabotage detection of whether to use them or not in future ones
< sipa>
i suspect that that behavior (halving CPU speed) will disappear before it becomes available in mainstream hardware
< luke-jr>
so just pretend the ones with it don't exist?
< yevaud>
I went to fix the display of -par=0 output printing nonsensically "using 0 threads for script verification" and looks like wumpus beat me to it. good stuff.
< yevaud>
I've been playing with bitcoind to see how small you can get the footprint of it for embedded into other applications. turns out very well. with no dbcache, no threads, no mempool, and no connections, it's almost invisible. slow, but that's not the goal.
< elichai2>
Is anyone having problems commenting on PRs? I'm trying to ACK something and I get `You can't comment at this time. `
< elichai2>
github is misbhaving :/ i'll throw this here for now in case I forget to comment when github is working properly again
< elichai2>
#18162 `Nice find! I'd refactor this out to a function that returns `Optional<struct tm>` but this can be done in a future PR ACK 6558ec35654d1a9990dcb534b18b5c88c5a4e165`
< gribble>
https://github.com/bitcoin/bitcoin/issues/18162 | util: Avoid potential uninitialized read in FormatISO8601DateTime(int64_t) by checking gmtime_s/gmtime_r return value by practicalswift . Pull Request #18162 . bitcoin/bitcoin . GitHub
< jonatack>
elichai2: after seeing that error, i refreshed the page and the comment was posted
< wumpus>
I had that once too, posting something, get an error, press 'send' again, error, again, error, then refresh the page and to my horror the comment is posted 10 times or so :-)
< emilengler>
sometimes trying stuff in the incognito mode of the browser works for some reasons
< ryanofsky>
i'm also getting errors commenting and even previewing comments, and I get a "Looks like something went wrong!" 500 error while trying to log in incognito