search

andytill / erlyberly

erlang tracing for the masses
https://twitter.com/erlyberlytips
GNU General Public License v3.0
693 stars 43 forks source link

process list stays empty after disconnect and reconnect #127

Open Licenser opened 8 years ago

Licenser commented 8 years ago

This can be reproduced by:

1) Start erlyberly and connect to a remote node => observe the process list to fill 2) use the disconnect button => observe the process list to become empty and the connection window show up 3) reconnect to this (or another node) => the process list remains empty

ruanpienaar commented 8 years ago

Any java crashes? Are you running the Jar in windows?

Licenser commented 8 years ago

That's a no on both accounts, OS X and no crashes in the shell.

ruanpienaar commented 8 years ago

Compiled from Master? I know we've not updated the 0.6.9 target on the Jar yet....

Licenser commented 8 years ago

Used this line from the readme:

git pull origin && ./mvnw clean compile install assembly:single && java -jar target/*runnable.jar

so yap that's master I guess :) admittedly my java is rustly with sprinkles of PTSD so I simply trust that this does the right thing :D

ruanpienaar commented 8 years ago

I'll try re-producing the same :)

Licenser commented 8 years ago

Good greif no, don't try to reproduce my java PTSD, if you're happy remain that way :D - just kidding know you're talking about the bug. it's a 100% hit for me so every time I get the same result.

ruanpienaar commented 8 years ago

Haha! I'll keep my java therapy-free. I'm coming up with a plan to run EB in a debug/verbose mode, to aid investigation.

@andytill- before i waste my time, is there a quick way to debug the proc classes?

andytill commented 8 years ago

@ruanpienaar repress that anger!

There is the debugger, that is about it. I have wanted to add debug logging for a while but don't want to add a dep to log4j. If you're going to log could you try out the logger in the standard library first? I have no idea if it is any good though.

I couldn't reproduce this bug. I remember a while back that it used to struggle with nodes that had over 100k processes, maybe the rpc call is timing out if that is the case. Riak has nowhere near that number of processes so I don't encounter that now.