14:01:19 <selsta> is there anything else we can try about this issue? https://github.com/tevador/RandomX/issues/262
14:01:54 <selsta> somehow I can't imagine that JIT is broken for a year now and no other program (e.g. browsers) run into this
14:02:36 <hyc> isn't it only broken with particular SDK ?
14:03:01 <selsta> it's weird, using an old SDK solves it when starting monerod from the command line
14:03:09 <selsta> but when I start monerod from the GUI as a detached process it also crashes
14:03:29 <hyc> then maybe it's an app signing issue or similar
14:04:01 <selsta> I had to add this workaround now https://github.com/monero-project/monero-gui/pull/4157
14:04:18 <hyc> ugh
14:05:14 <hyc> the fact that it works from CLI tends to imply that the code is fine as is, but maybe some other metadata is missing
14:05:21 <hyc> like app signature crap
14:06:02 <hyc> what about running randomx-test or randomx-benchmark ?
14:06:43 <selsta> will try to run randomx-test
14:07:24 <hyc> we ought to be able to see some API call failing, returning an error code
14:07:39 <hyc> I guess there's no strace on macos tho. maybe dtrace or something similar
14:09:48 <selsta> this is where it crashes when starting as a detached process https://www.irccloud.com/pastebin/zm47Sarz/
14:11:47 <hyc> that doesn't tell me much, location-wise. it just says something in monerod.
14:11:55 <hyc> and it says a NULL pointer deref
14:12:21 <hyc> is that a binary built with debug symbols?
14:14:16 <selsta> it's the release binary we built using depends/gitian
14:14:29 <selsta> that uses the old sdk
14:14:52 <hyc> then not a debug build.
14:16:48 <hyc> threads 19 and 20 also have no valid stack traces
14:16:53 <hyc> odd
14:17:06 <selsta> i think the crash itself isn't interesting, it fails to calculate correct hashes so once that's fixed the crash should also be gone
14:17:25 <hyc> shouldn't crash on invalid hash results
14:17:51 <hyc> anyway... no luck with randomx-test?
14:20:19 <selsta> working on it
14:20:26 <hyc> I'm still on MacOS12
14:20:30 <selsta> i'll use xmrig benchmark because that one also crashes
14:20:32 <hyc> no problems here
14:20:45 <selsta> yea, it only happened with macOS 13
14:21:03 <hyc> if you can gt a debug build and run it under their debugger maybe we can get more useful info out
14:21:52 <hyc> also I've been considering installing Asahi Linux and leaving macOS behind completely
14:29:12 <selsta> interesting so xmrig benchmark doesn't crash
14:29:16 <selsta> trying randomx-tests now
14:31:09 <hyc> interesting - Intel giving up on bitcoin ASICs https://blockchainbeat.co/chip-giant-intel-abandons-bitcoin-asic-production-bitcoin-news/
14:31:50 <selsta> randomx-tests also works, so it's only monerod somehow
14:32:02 <hyc> that's very strange
14:32:15 <hyc> wait that's all from CLI then?
14:32:23 <ofrnxmr[m]> hyc: Halving is next year? And those tx fees from nfts are paying the bills
14:32:26 <selsta> no, starting from the GUI as detachedProcess
14:32:27 <ofrnxmr[m]> Arent*
14:32:35 <hyc> selsta: ok, cool
14:32:39 <selsta> monerod from CLI works without issues
14:33:16 <hyc> then is it even related to randomx at all?
14:33:29 <hyc> maybe it's some missing environment var or something silly
14:33:46 <selsta> it's solved when disabling JIT so not sure
14:35:15 <selsta> i'll try around with env vars and working directory
14:36:51 <hyc> have we tried this before - comment out the #define USE_PTHREAD_JIT_WP in randomx/src/virtual_memory.c ?
14:39:41 <selsta> same
14:41:26 <selsta> https://stackoverflow.com/questions/75327638/sys-icache-invalidate-slow-on-m1
14:41:46 <selsta> for example this user seems to be using macOS 13 and while they say it's slow it seems it's at least working for them
14:42:58 <sech1> If cache invalidate is not working properly, it can give all sorts of random results, i.e. CLI working and GUI not working
15:10:56 <hyc> unfortunately cache invalidate is an OS function, not something we can fix if it's broken
15:11:16 <hyc> although I suppose we could duplicate its functionality ourselves in asm code
15:24:40 <selsta> https://github.com/apple/darwin-libplatform/blob/main/src/cachecontrol/arm64/cache.s
15:37:10 <hyc> strange, that page says latest commit Nov 18 2020 but if you select History it shows commits up to 2021
15:44:41 <DataHoarder> authored on vs commited on