-
sech1
selsta moneromooo wfaressuissia I got this crash again:
paste.debian.net/hidden/6b06edc2
-
sech1
anything else I can do in gdb before I kill the process? I'll wait for a few hours.
-
selsta
it seems to be the race fixed in
monero-project/monero #7873
-
selsta
I think you can close it and recompile latest release-v0.17
-
selsta
still don't know why it gets triggered so frequently on your machine
-
sech1
Now I need to understand how the fix works
-
sech1
I mean, it removed "std::vector<cryptonote::blobdata> fluff_txs;" alltogether, but where it was used?
-
selsta
btw you can use gcore inside gdb to save the core file so that you don't have to keep it open
-
selsta
though don't remember the exact command
-
sech1
so the fix moved "fluff_txs" from p2p connection to a separate struct, ok
-
sech1
it crashes so often probably because of slow HDD + purged node. It was a full node before but it didn't crash.
-
sech1
maybe also because p2pool connects to it and ZMQ is activated
-
sech1
the difference with my other stable node is HDD/SSD and purged/full
-
sech1
building latest release-v0.17, I'll run it under gdb for a few days too
-
sech1
to make sure that the fix actually fixes the crash. My server is a perfect "Canary" :D
-
wfaressuissia
`node_server: add race condition demo` all these demos are reproducing with almost 100% probability fixed problem for debug/release build
-
wfaressuissia
so anyone can run that demo under gdb without fix to know what was the problem
-
monerobull[m]
17.3 is getting prepared, i assume we're getting it 1st of November, right?
-
selsta
luigi1111: merges today?
-
luigi1111
yes
-
fluffypony
2021-10-25 15:49:22.034 I SYNCHRONIZED OK
-
fluffypony
free(): corrupted unsorted chunks
-
fluffypony
Aborted
-
fluffypony
just had a long-ish running Linux node die
-
fluffypony
moneromooo: something you're aware of or should I see if it dumped?
-
wfaressuissia
`free(): corrupted unsorted chunks` very interesting, any info ?
-
fluffypony
this is on v0.17.2.3-release btw
-
wfaressuissia
anything in dmesg /journalctl ?
-
fluffypony
it was running in a screen session, so nothing in journalctl
-
fluffypony
dmesg looks normal too
-
wfaressuissia
how screen may hide abnormally terminated processes from journalctl ? gdb - yes, screen/tmux - likely no
-
wfaressuissia
"dmesg looks normal too" it's probably normal, so no info without coredump then
-
moneromooo
That's a bad bug and needs a fix.
-
fluffypony
bah - no coredumps enabled
-
fluffypony
wfaressuissia: sorry I saw journalctl and read systemd, my brain is on the fritz today
-
fluffypony
journalctl looks clean too
-
fluffypony
I've turned dumps on
-
fluffypony
will see if it happens again
-
hendursaga
Fingerprint is still 81AC 591F E9C4 B65C 5806 AFC3 F0AF 4D46 2A0B DF92 right?
-
moneromooo
Part of your message is missing (the part where you say what that fingerprint is meant be for).
-
luigi1112
.merges
-
xmr-pr
7995 7996 8002 8003 8004 8006 8007 8011 8018 8019 8022