00:32:43 I've recently tried running monerod on alpine linux, which is a musl based distribution, and the performance are catastrophic to say the least. Initial fast-sync is two times slower than normal and once checkpoints are passed, I estimated it would need 62 hours to process 150000 blocks (95%). Some better C++ developer could answer the question, but would it be easy to make monerod 00:32:45 compile with jemalloc on musl targets ? This is something that [ripgrep had to do for musl system](https://andygrove.io/2020/05/why-musl-extremely-slow/) 03:06:00 My Alpine builds just segfaulted D: 09:42:03 weird 09:42:34 can ypu do perf? 10:40:02 Mamma Mia! 11:09:40 Ok I've increased log-level and my node is spamming messages: 11:09:41 ``` 11:09:43 2024-10-09 HH:MM:SS.xxx I [IP:PORT OUT] Failed to lock m_sync_lock, going back to download 11:09:45 ``` 11:10:39 That is normal. 11:11:03 Alright 11:11:11 It means a thread is looking "is anything verifying blocks atm ? Yes, so I won't and will download more stuff instead". 17:01:01 hello 18:38:05 spirobel, Czarek Nakamoto is the author of monero_c 18:47:59 SNeedlewoods: I'm back if you need something. 18:51:00 Still looking through your code, my code and the current state and taking notes to get a clear overview first 18:59:14 Update on my alpine story. I've managed (with pain and suffering) to compile master branch on Alpine in release mode + no stack_trace + 0xfffc `new-miniupnpc` PR + easylogging patch. It now dynamically link to musl which permit me to override default malloc. It's now like 4x time faster. 18:59:58 A shame you cannot override C++ default `std::alocator` that would have simply things a lot.