00:49:05 I did the very professional technique of starting my node and then typing the command in really fast 01:37:22 i was/am having a lot of "no connection" log spam, so i set out and in peers to 0, and checked printcn -> crash :D 03:27:01 hello all, weird issue when trying to build monero with a custom ARCH setting (cmake var that controls -march= and related feature flags): 03:27:01 1) by default, i.e. `make` with no arguments, i end up with `ARCH=native` in the CMakeCache, -march=native passed to compiler as expected, and successful build 03:27:03 2) forcibly setting `cmake -DARCH=default` doesn't pass -march at all, instead passing -maes, also as expected, and also succeeds 03:27:05 3) setting `cmake -DARCH=sandybridge` passes -march=sandybridge, as expected, but linking fails to find many (probably ANY) RandomX and CryptoNight related symbols -- full 635-line output at https://paste.debian.net/1398570/ 03:27:07 anyone have any ideas why this might be happening? 03:28:44 hello all, weird issue when trying to build monero with a custom ARCH setting (cmake var that controls -march= and related feature flags): 03:28:45 1. by default, i.e. `make` with no arguments, i end up with `ARCH=native` in the CMakeCache, -march=native passed to compiler as expected, and successful build 03:28:47 2. forcibly setting `cmake -DARCH=default` doesn't pass -march at all, instead passing -maes, also as expected, and also succeeds 03:28:49 3. setting `cmake -DARCH=sandybridge` passes -march=sandybridge, as expected, but linking fails to find many (probably ANY) RandomX and CryptoNight related symbols -- full 635-line output at https://paste.debian.net/1398570/ 03:28:51 anyone have any ideas why this might be happening? 03:28:53 ETA: this is tag v0.18.4.2 on debian bookworm, so gcc 12.2.0 04:34:31 hmm so it turns out that `-march=sandybridge` does not define __AES__. it does with `-maes`, and that is also being passed, so it seems like it should still work but i guess something is getting lost somewhere 04:35:05 hmm so it turns out that `-march=sandybridge` does not define `__AES__`. it does with `-maes`, and that is also being passed, so it seems like it should still work but i guess something is getting lost somewhere 05:34:04 whew. solved, lol. turns out that setting ARCH=sandybridge *also* sets ARCH_ID to the same value, but the latter should be x86_64, not sandybridge. so `cmake -DARCH=sandybridge -DARCH_ID=x86_64` fixes everything