.merges

Merge queue empty

.merge+ #2933

...

.merges

.merge+ #2959

Added

I noticed that DaemonManager does not trigger signals because monerod is started with startDetached().

Output logging from the daemon to the Settings -> Log page is broken.

And stateChanged is also never executed. This causes an issue where if monerod crashes during startup it takes ~2 minutes before the wallet gives up trying to connect to the daemon.

Ideally the wallet detects that monerod has crashed (by checking for exitcode != 0) and show the daemonStartFailure dialog with an error message. This can be done by connecting to the finished signal.

But, for signals to work the daemon has to be started with start() and the --detach flag removed.

That means that the daemon will stop when the wallet is exited, so the prompt that asks to keep monerod running will break.

A workaround is to stop monerod and then start it again detached when the wallet is closed.

I'm working on a patch to address all of this.

Not sure this is a good way

And the workaround idea doesnt' sound good to me

"I'm working on a patch to address all of this." -> please consider to submit multiple separate PRs

What would be a better way?

"please consider to submit multiple separate PRs" -> Sure

Regarding the daemon - i see absolutelu no reaso to start it detached

Just start it always attached

And the detach when necessary

So remove the functionality to keep it running on wallet close?

No

Just detach from it

I looked for that, QT doesn't provide an API afaik

xiphon: lists.qt-project.org/pipermail/development/2019-February/035333.html

So it appears that either we don't have output logging, detecting crashes or state changes. Or we can't keep the daemon running on exit (without a workaround).

Imo it would be very helpful to show why monerod failed to start.

Just yesterday I spent 2 hours debugging with someone to figure out why the daemon wouldn't start in their GUI.

It's hard to tell if it's a corrupted blockchain, incorrect daemon flags, AV killing the process, or as it turns out, an issue in pop_block() without output logs or error messages.

Why not instruct user to start monerod manually if they want to keep it running or apply ugly workaround?

"It's hard to tell if it's a corrupted blockchain, incorrect daemon flags, AV killing the process, or as it turns out, an issue in pop_block() without output logs or error messages." -> all this should be in monerod logs

except "AV killing the process"

^ but "output logs or error messages" won't help you either in that case

"all this should be in monerod logs" -> The user isn't going to look for monerod logs (or even know that they exist) if they don't know that monerod crashed. Right now when monerod crashes they get nothing. It hangs for two minutes and then the status says "Disconnected". Not very helpful and not very actionable.

"but "output logs or error messages" won't help you either in that case" -> probably exits with a return code that is not 0 or 1, and the absence of an error message can give it away

.merges

.merge+ #2937 #2947 #2950 #2961 #2962

Added