Process Lasso 5.1.0.72 Bug

Started by Miroku4444, April 16, 2012, 07:17:27 PM

Previous topic - Next topic

Miroku4444

I'm experiencing a problem with the 5.1.0.72 update. When i go and shut down my PC, its starts to shut down and then stops. I then have to hit turn off again, and then it shuts down. Thru a bunch of digging, i discovered it was Process Lasso causing it. If i shut down Process Lasso before i turn off my PC, everything shuts down like normal. This started with version 5.1.0.72.

WinXPSP3

edkiefer

now that you mentioned it, this happened to me last night to shutting down but I didn't look to hard and just shutdown again (that was with 5.1.0.70) .
I just tried a reboot and what happens on first try is PL shuts down but windows is still up .
Bitsum QA Engineer

Jeremy Collake

Thanks for the report. After some confusion, I believe I may see a problem. During the back-port there may have been an issue. I am evaluating it now and will issue an immediate fix if that's what it is.
Software Engineer. Bitsum LLC.

edkiefer

I am running 5.1.0.72 (updated this morning) and I noticed it does shutdown a few services along with PL but that is it . I have to go through same shutdown second time an works .

For me I don't have PL as service, I manually start it each day/startup .
Bitsum QA Engineer

Jeremy Collake

Yes, I have determined the cause of the problem and am fixing it in v5.1.0.74. Very sorry. Process Lasso is basically, inadvertently, telling Windows 'NO, I don't want to shut down'.
Software Engineer. Bitsum LLC.

Jeremy Collake

This update will be available in about 30 minutes. Note that this issue *only* affects Windows XP and Windows 2000. I am again very sorry for the problem. Until then, you may simply exit Process Lasso before rebooting, or use an older version from the Support area.
Software Engineer. Bitsum LLC.

Miroku4444

Thanks for the fast fix, much appreciated.

Jeremy Collake

#7
I am uploading v5.1.0.74 now. You may confirm its fix at your leisure. I confirmed it in XP this time, so all should be good. I am concerned about a subset of users who saw the original issue in .68 (not this issue), and am therefore engaged in some refactoring to make sure nobody ever sees that older issue again. It was separate from this one though. This issue prevented shutdown. The older issue simply showed a non-important error message during shutdowns that took too long.
Software Engineer. Bitsum LLC.

edkiefer

Bitsum QA Engineer

Miroku4444

Its fixed with me also, thanks.

Jeremy Collake

Software Engineer. Bitsum LLC.

Tarnak

I am still getting this at shutdown/reboot in XP3 with v5.1.0.74,  in one of my snapshots:

processgovernor.exe - DLL Initialization Failed

"The application failed  to initialize because the window station is shutting down"

together with the sound referred to in my post in the thread - Errors when logging out on notebook

Jeremy Collake

#12
Thanks. Yes, I expected such. I feared those who saw the original condition might see it from time to time. Take no action and it should shut down fine. The refactored code is being tested now. Please understand, my first priority was an immediate fix to a problem that made things worse (from a few users seeing an insignificant message box to a lot of users not being able to shut down). Being in the middle of my taxes, I had to issue a quick fix. The next fix will be the most tested, proper one, that addresses the underlying issue you see (the governor's new self-recovery mechanism kicking in because the time difference between shutdown notification to the GUI and shutdown notification to the governor exceeds 10 seconds), a rare scenario. I did improve things a bit by making sure the self-recovery mechanism turned off as soon as WM_QUERYENDSESSSION was issued to the GUI, but still this did not cover *any* time spread, which was the underlying issue. In Windows Vista and above the shutdown messages are sent differently, and therefore this is not an issue. In XP this is only an issue on systems that are pretty slow to shutdown.

The sound is most likely the message box itself, being a system error message, it plays the system error sound.

Recommended action is simply ignore the message box (don't click it). The sound is the message box trying to display or re-display itself.

Final fix to that older, less serious issue seen by a subset of 2K/XP users will come very soon. I will just be darned if I'll rush it again ;).

My quote above:
Quote
I am concerned about a subset of users who saw the original issue in .68 (not this issue), and am therefore engaged in some refactoring to make sure nobody ever sees that older issue again.
Software Engineer. Bitsum LLC.

Jeremy Collake

#13
Ok, in v5.1.0.76, the explicit shutdown order was changed so that the GUI will always terminate before the governor. This will fix the issue without any time bias (shutdown take 3 weeks? That's fine).

Again, I do apologize for not doing this first. I had hoped that the receipt of WM_QUERYENDSESSION by the GUI would always pre-empt the receipt of WM_ENDSESSION by the governor, regardless of shutdown order, but was apparently incorrrect in this assumption (which I was concerned about). I was attempting to make as few changes as possible, instead of taking a broader look at the issue.. I suppose.

The original issue, of course, was me backporting the governor recovery mechanism to v5.1.0.68. That was the true mistake here.

UPDATE: Now issued, all tests have been fine, I'm pretty darn sure this will go good for *every user*.
Software Engineer. Bitsum LLC.

Tarnak

Quote from: bitsum.support on April 20, 2012, 11:45:59 AM
Ok, in v5.1.0.76, the explicit shutdown order was changed so that the GUI will always terminate before the governor. This will fix the issue without any time bias (shutdown take 3 weeks? That's fine).

I must have a slow system. ;)

Quote from: bitsum.support on April 20, 2012, 11:45:59 AM
Again, I do apologize for not doing this first. I had hoped that the receipt of WM_QUERYENDSESSION by the GUI would always pre-empt the receipt of WM_ENDSESSION by the governor, regardless of shutdown order, but was apparently incorrrect in this assumption (which I was concerned about). I was attempting to make as few changes as possible, instead of taking a broader look at the issue.. I suppose.

UPDATE: Now issued, all tests have been fine, I'm pretty darn sure this will go good for *every user*.

..."the receipt of WM_QUERYENDSESSION"  ..."WM_ENDSESSION"   .... over my head, I'm afraid! ;)

Quote from: bitsum.support on April 20, 2012, 11:45:59 AM
The original issue, of course, was me backporting the governor recovery mechanism to v5.1.0.68. That was the true mistake here.

UPDATE: Now issued, all tests have been fine, I'm pretty darn sure this will go good for *every user*.


This has fixed it for me. :)

Jeremy Collake

Thanks Tarnak. Although I was/am *very* confident in this final fix, I must admit some trepidation in reading your update, lol.

Your system isn't necessarily slow to shutdown, I shouldn't have implied that. It's just a certain point, the closure of applications, the shutdown process was slow for whatever reason. Maybe an open document in another application, or a service that refused to terminate. Who knows. However, NOW, it doesn't matter, as the ordering has changed so that this issue is null and void. The GUI always terminates before the governor on all systems, all configurations. Thus, the governor recovery mechanism that resides in the GUI (at present at least) is not a player.

This has actually helped with version 6, as it as an issue I would have encountered there too. So, I look at it as ironing out bugs to come ;). Now that things are returning to normal after what I call 'hell' (April - tax season), I am quickly getting back on track here and expect version 6 very soon.
Software Engineer. Bitsum LLC.