Hi old in SETI but new on this boards so please forgive me if i post this on a wrong thread.

Using the unreleased 7.16.1 under Linux but i encounter a weird situation.
First some explanations of the problem:
I start the boinc client on one terminal and the manager on another.
Works normally and i could see all working even with the manager GUI.
My manager is set to NOT run the client on start up and enable the manager exit dialog.
If i close the manager using the file option it close as expected leaving the client running on the other terminal.
But if i close the Manager GUI using the red (x) bottom if terminates the manager as expected but exit the client too that is not expected.
That not happening on the old 7.14.2 manager.
Any suggestion?

<edit> That happening randomly, some times it works OK and sometimes no, i only notice if that error happening when i see the crunching stopping or see the temps down telling no crunching is active on the host. Something very annoying BTW.

Haven't noticed this. Certainly exiting manager when I have work by clicking the x in top RH corner results in client continuing to run. And being out of work on one machine a the moment, I have been able to try this and client has continued to run even with no tasks.

I rolled my own a few weeks ago so there may have been changes between when I got mine from git-hub. Don't know where yours came from so it could be an earlier or later build.

I rolled my own a few weeks ago so there may have been changes between when I got mine from git-hub. Don't know where yours came from so it could be an earlier or later build.

There have been no effective updates to the v7.16.1 branch since it was made on 30 July. But there is one horrible bug in the Manager code which you need to correct if you built that as well.

It's an easy fix - remove one stray comma from a library file - but the side effects are critical. See #3247, #3249

I fond a way to replicate the problem.
That happening the first time you exit the manager.
Then if you go to the terminal who has the boinc client running (now stopped) and restart it, the next time you close the manager it works as expected leaving the client running.
That only happening if you close with the red X bottom, not when you shutdown using the file exit tab.
My code has the path for the bug Richard talk about.

Thanks. I can try to replicate tomorrow, and see if I can find anything in the code that would explain it. And if so, I can report it upstairs.

Just before I try to replicate, do you use any command line parameters on either the client or the manager launches?

Thanks. I can try to replicate tomorrow, and see if I can find anything in the code that would explain it. And if so, I can report it upstairs.

Just before I try to replicate, do you use any command line parameters on either the client or the manager launches?

No simply: ./boinc & ./boincmgr one on each terminal and nothing else.

The boincmgr is set to not start the client and not stop it at exit.

That's odd: after cleaning up the duplicates, I count six visible posts before this one. But the index said two before I came back for this comment.

There have been no effective updates to the v7.16.1 branch since it was made on 30 July. But there is one horrible bug in the Manager code which you need to correct if you built that as well.

Thanks Richard,
I found the offending line on one of my two Linux machines and have manually reset the value to 0. As I don't exit BOINC for weeks at a time and now have cc_config.xml the way I like it, it shouldn't be a problem for me but will do a fresh build when one with fix becomes available.

The fix is trivial, and you can do it on your own machine in seconds.

Remove the comma from the end of https://github.com/BOINC/boinc/blob/client_release/7/7.16/lib/cc_config.cpp#L506

That's -

Folder:	lib
File:	cc_config.cpp
Line:	506

which currently reads

"        <allow_gui_rpc_get>%d</allow_gui_rpc_get>\n",

I'm trying to play this by the book, but last fortnight I had my editor loaded with the offending line during the developer conference call. I'll do the same tonight, and if the offender doesn't pull his finger out and fix it (as he's been told to), I will break protocol and fix it myself. That may result in the removal of my privileges to edit code, but they will know what they are doing.

and if the offender doesn't pull his finger out and fix it

Because I wasn't sure what happens when you fiddle with things I downloaded a fresh copy of boinc-master.zip and the offending line seems to have been fixed.

Edit: I inadvertently downloaded 7.15 so perhaps not.

and if the offender doesn't pull his finger out and fix it

Because I wasn't sure what happens when you fiddle with things I downloaded a fresh copy of boinc-master.zip and the offending line seems to have been fixed.

Yes.

Master good, release branch bad. That's my main grouse: the bug was fixed on 13 August and merged into master the same day (#3249). It should have been copied into the release branch within hours of the patch being tested and approved - which you can see I did, again on the same day. You can also see that LoctutusOfBorg and Germano updated their releases within three days.

Yet here we are, over a month later, and it still hasn't been done. Protocol says the Release Manager should do it: I did that role once, so I know the form. And I know it's been disregarded. See https://github.com/BOINC/boinc-policy/blob/master/Development_Documents/Client_Release_Process.md, especially the 'Process' section and the 'Prepare for a Release' and 'Release Branch' sub-sections.

Back to your home build: if you simply took that one line, or even the whole file, from a 'master' download, you'll be fine. If you did the whole build from master, you run the risk of some other bug, as yet undiscovered, catching you and biting you on the ankles. See, for example, #3292: it took David almost a fortnight to notice and correct that one, but at least it was caught before it left master.

Edit: I inadvertently downloaded 7.15 so perhaps not.

'master' is always indicated as a development version by the odd number in the middle position. It should have been revised to v7.17 by now, to indicate the new post-v7.16 cycle, but that hasn't been done either.

Well, the right questions were asked and acknowledged, so I didn't have to go nuclear. We might see some progress towards v7.16.2 soon.

Cool.

Thanks for keeping us informed on this. I only know of two others on the cpdn boards who roll their own but that is enough for me to have posted the link to this conversation.

Github has been updated overnight to v7.16.2

That includes the fix we're talking about here, plus at least 60 others. There is currently a delay in updating the translations, so it isn't yet ready to begin full release testing - but English speakers with a sense of adventure are welcome to try rolling their own. Caveat emptor.

Github has been updated overnight to v7.16.2

And now I have lots of pretty colours scrolling up my terminii.

Edit: And confusingly the zip is named boinc-client_release-7.7.16.zip

Any news? Closing with the red X on the manager still makes the client exit on the 7.16.2 even when th emanager is configurate dto not stop the client on exit.