Ctrl-R not working?

[georgemk]

@betterred
The keystrokes are seen by Calibre and actioned. It is the result that differs. In normal mode, Calibre doesn't close down fast enough for the restart to work as the restart process sees the original process and exits.

[BetterRed]

But why would calibre behave differently in Safe and Normal modes - would it even know. More likely that something which runs in Normal mode but not in Safe mode is 'interfering' in Normal mode.

Eyeballing the two process list (Safe and Normal) might provide a clue, or do a diff if you don't trust your eyes

BR

[compurandom]

...and now I have to eat my words.

Running calibre 5.3 (linux), just updated a plugin and installed a second plugin, then hit the restart button.

Calibre closed its window, waited a few seconds, and popped up the "another calibre is running" dialog. I waited a few more seconds and manually re-ran it and all is good.

So restart is still giving me problems, even when it's highly unlikely to be thrashing.

Irritating, but not a big deal.

[georgemk]

Originally Posted by compurandom

...and now I have to eat my words.

Well interesting. So it is happening on Linux too.

I've given up as I've spent probably 24 hours doing various tests and code reading over the last week.

I know how to work around it. So I'll live with the work round.

[DNSB]

Originally Posted by compurandom

...and now I have to eat my words.

Running calibre 5.3 (linux), just updated a plugin and installed a second plugin, then hit the restart button.

Calibre closed its window, waited a few seconds, and popped up the "another calibre is running" dialog. I waited a few more seconds and manually re-ran it and all is good.

So restart is still giving me problems, even when it's highly unlikely to be thrashing.

Irritating, but not a big deal.

If you remove the updated plugin and the new plugin, does the problem go away? If so, can you share the names of those two plugins?

[compurandom]

Originally Posted by DNSB

If you remove the updated plugin and the new plugin, does the problem go away? If so, can you share the names of those two plugins?

This is NOT a new problem, and highly unlikely that it is related to any specific plugin being installed. I just didn't expect it with my recent memory upgrade -- it was only a few seconds, where it use to be 10 seconds, so maybe the memory upgrade helped. It doesn't happen every time. I've seen this through all of version 4 and probably before that.

I just ran restart from the favorites menu (without adjusting any plugins) and it worked fine.
I'd bet that if I removed a plugin, it would then fail. Or maybe not.
Or maybe it fails right after it does a news download or count pages and there's a lingering inactive background thread.

If I watch the calibre processes, I can see the main thread exit and then the worker threads exit, and calibre won't restart until the last worker thread is gone.

I'm pretty sure this is related to idle background threads that take a while to wake up and exit, and it only occurs when there *is* a worker thread, and system load characteristics make it exit late.

Maybe the problem isn't there in safe mode not because of safe mode, but because calibre was freshly started and didn't get a chance to start any worker threads. Or maybe the worker threads were sill in memory instead of VM and took no time to exit.

I think the only way to fix this bug is to figure out what the blocker is, and just loop on waiting for it to shutdown, maybe with a nice UI countdown timer, and THEN restart. The end result would be the same time with less user interaction. And for some, less frustration.

I bet people with SSD disks never see this, but I was wrong about the memory upgrade fixing it.

[ownedbycats]

Another thing that might be in factor is that Windows 10, at least, uses memory compression by default. It's definitely more efficient than paging it out but I wonder if it may cause issues if Calibre gets compressed.

In the case of memory compression, its the CPU that's the bottleneck rather than the HDD.

[georgemk]

Originally Posted by compurandom

I bet people with SSD disks never see this, but I was wrong about the memory upgrade fixing it.

Guess who only has SDD drives in his machine

[kovidgoyal]

@georgemk: See if this helps: https://github.com/kovidgoyal/calibr...d7f91aa13cb001

Now calibre will start even if the listener could not be established, you will just get an error message that you can close/ignore.

[compurandom]

Originally Posted by ownedbycats

Another thing that might be in factor is that
In the case of memory compression, its the CPU that's the bottleneck rather than the HDD.

Good point. And I do have memory compression enabled.
Perhaps an easy experiment is on order.

[georgemk]

Originally Posted by kovidgoyal

@georgemk: See if this helps: https://github.com/kovidgoyal/calibr...d7f91aa13cb001

Now calibre will start even if the listener could not be established, you will just get an error message that you can close/ignore.

You are a genius sir. Yes it works for Ctrl R and Ctrl Shift R. And for completeness when I updated a plugin, the button request restart work as well. Ditto the Preference menu debug menu option.

Thank you so much.