libera #monero-dev

5 May 2023

Official repo: github.com/monero-project/monero || Don’t ask to ask, just ask:) || Monero GUI dev is at #monero-gui || Monero website dev is at #monero-site || Non-dev questions -> #monero, please || No fork help, you’re on your own

• 13:22
  dr00tb_
  nick /dr00tb
• 17:36
  john_alan_
  hi folks - I'm seeing a segmentation fault around 99% complete sync on ARM64 macOS/Darwin
• 17:37
  john_alan_
  I've tried on multiple occasions, nothing obvious in logs. On 0.18.2.2
• 17:37
  john_alan_
  anything reported/known, before I make a GH issue?
• 17:38
  john_alan_
  verbose logging simply shows: ```2023-05-04 09:25:36.240 T Blockchain::have_block_unlocked
• 17:38
  john_alan_
  2023-05-04 09:25:36.240 T BlockchainLMDB::block_exists
• 17:38
  john_alan_
  2023-05-04 09:25:36.240 T BlockchainLMDB::block_rtxn_start
• 17:38
  john_alan_
  2023-05-04 09:25:36.240 T Block with hash c28336da72d1abc42b5c4adb785d8dc0debc779a614a86ef8d43305000b5eea0 not found in db
• 17:38
  john_alan_
  2023-05-04 09:25:36.240 T mdb_txn_safe: destructor
• 17:38
  john_alan_
  2023-05-04 09:25:36.240 T BlockchainLMDB:: get_alt_block
• 17:38
  john_alan_
  2023-05-04 09:25:36.240 T BlockchainLMDB::block_rtxn_start
• 17:38
  john_alan_
  2023-05-04 09:25:36.240 T mdb_txn_safe: destructor```
• 17:39
  john_alan_
  when I say I've tried on 'multiple occasions' I mean, I've resyncd 3 times.
• 17:54
  moneromooo
  Get a stack trace with gdb.
• 17:54
  moneromooo
  ie, gdb /path/to/monerod path/to/core ; bt
• 17:55
  moneromooo
  Or coredumpctl debug if you have systemd
• 17:55
  selsta
  john_alan_: self compiled?
• 17:56
  selsta
  or how did you install it?
• 17:57
  selsta
  I assume this issue will be relevant monero-project/monero #8657
• 17:57
  john_alan_
  hey both
• 17:57
  john_alan_
  installed via home-brew
• 17:57
  john_alan_
  (on macOS), happy to help debug, checking out 8657 now...
• 17:57
  selsta
  you have to use the release binaries from getmonero.org
• 17:58
  selsta
  or you have to set `MONERO_RANDOMX_UMASK=8` env var, both work
• 17:59
  john_alan_
  ahhhh
• 17:59
  john_alan_
  so home-brew default is broken?
• 17:59
  selsta
  it's not homebrew's fault and it's unclear if this is a macos bug or not
• 18:00
  john_alan_
  this is a bottled version I'm using form homebrew i.e. not compiled locally, we could add a fix to their build script
• 18:00
  selsta
  moneromooo: does this look like a serialization issue on 32bit? monero-project/monero #8837#issuecomment-1536588024
• 18:01
  john_alan_
  happy to file with homebrew if not already done, either way, thanks both.
• 18:02
  selsta
  john_alan_: I'm didn't file any issue yet for it, I feel like they will say we should fix it on our side
• 18:03
  john_alan_
  ya I get you, arm64 bin from getmonero is working
• 18:03
  john_alan_
  how weird
• 18:03
  john_alan_
  I wonder what the root cause is
• 18:05
  john_alan_
  I guess some weird JIT security thing
• 18:05
  selsta
  RandomX cache invalidation somehow isn't working
• 18:05
  john_alan_
  I remember macOS treats JIT differently to BSD
• 18:05
  selsta
  but this whole thing only started happening on macOS 13 so we assumed it's a macOS bug
• 18:05
  john_alan_
  macOS security issue?
• 18:06
  john_alan_
  related to JITc tho right?
• 18:06
  selsta
  I don't think it's a security issue, just JIT cache invalidation isn't working causing it to calculate wrong hashes
• 18:06
  selsta
  MONERO_RANDOMX_UMASK=8 disables JIT
• 18:06
  john_alan_
  thanks selsta, does that impact performance?
• 18:06
  selsta
  yes
• 18:07
  john_alan_
  when I mean security, I mean macOS is being weird on security causing JIT to behave funny
• 18:07
  john_alan_
  they keep messing with JIT rules
• 18:08
  selsta
  there's this document and we do everything that's written there developer.apple.com/documentation/a…-in-time-compilers-to-apple-silicon
• 18:19
  john_alan_
  Thanks selsta
• 20:54
  moneromooo
  selsta: probably
• 20:54
  moneromooo
  Something reading unaligned words (I assume their arch doesn't like that, which is usually what bus errors means).
• 21:15
  john_alan_
  @selsta: does disable randomX JIT have any other impact outside of performance?