v3.99.38 Crash On Start-Up

Started by wineaux, October 17, 2010, 05:00:35 PM

Previous topic - Next topic

wineaux

Just downloaded v3.99.38 and am getting a crash on start-up that I cannot get to go away.

Windows 64-Bit Home Premium (All current patches)
ASUS Core i7 laptop
4GB RAM

Upon running the program you get: "An exception occurred within Process Lasso."  The program locks up and this point and must be shut down.

Jeremy Collake

Thanks for reporting that. I am looking into it immediately to see if I can figure out what might be going on. It doesn't seem to manifest in my early tests, but that doesn't mean much ;o.
Software Engineer. Bitsum LLC.

Jeremy Collake

Any seconds on this? I still haven't seen it in test beds.
Software Engineer. Bitsum LLC.

parkd1

I got this error too on windows start up and using the same version. I am using the 32 bit version. Keep up the great work on this great program. Are you able to test the program with the beta version that is out of SP1 for Windows 7?

Jeremy Collake

I think I know the cause here guys, and am fixing it. In the meantime, it *may* help if you reinstall and select 'Run with elevate rights' in the second of the InstallHelper dialogs. Then again it may not. Few things changed from .37 to .38, so I will track it down regardless.

I DO have access to the SP1 beta - being an MSDN subscriber - and will be testing with it, yes.
Software Engineer. Bitsum LLC.

Jeremy Collake

Actually, I'm not sure my previous guess will make a difference at all... I'm working on this, and out of curiosity, what languages are you guys using?
Software Engineer. Bitsum LLC.

wineaux


Jeremy Collake

Ok, one factor ruled out for sure ;). I'll likely run into this as I work tonight, but did v3.99.37 work? Do you remember the last version you used?
Software Engineer. Bitsum LLC.

Jeremy Collake

I dunno guys, I have a suspected cause, but am going to have to work on this more. You may try v3.99.39 beta, but I doubt it changes anything. This is a high priority issue, but strange I'm not seeing it. I'll have to test in some fresh test beds, maybe I'll see it then. Thanks for reporting, and any clues are appreciated. Regardless, I'll get it figured out.
Software Engineer. Bitsum LLC.

parkd1

English (USA)

v3.99.37 worked great. I just started the computer up and it did not crash so thinking this one might be random. I will try out the new build and see if it is still there.

wineaux

Previous version worked like a champ.  I'll try the new version as well and see if that helps.

wineaux

3.99.39 results are the same as 3.99.38.  Instant crash on start-up.  I'll keep checking for further updates.

Jeremy Collake

#12
I am pretty certain I know what code is causing the crash, but I don't know why. Anyone up to try a sample build to confirm this for me? ;)

I had added some code so that the GUI waits for the explorer.exe (the shell) to start up before it goes and tries to 'talk to it' in the form of creating a system tray icon. This seems to correlate with when you guys started seeing problems. The idea was to handle rare cases where someone actually had the GUI launched before the shell -- Yes, a rare condition indeed.

I have also noticed a typo on the critical section spin count that could have caused rare performance problems - this appeared maybe about that same time, but shouldn't ever have caused a crash, just bad performance of Process Lasso itself under some rare cases on certain systems. I've fixed that now.

This is my TOP priority right now, as it obviously doesn't leave a good impression with the user.

Here are sample daily builds with both these things changed:

http://www.bitsum.com/files/daily/prolasso.zip
http://www.bitsum.com/files/daily/prolasso64.zip

Software Engineer. Bitsum LLC.

wineaux

Installing now, but...

I am on a 64-bit version on Win-7 and have been using the 64-bit version of the program.  I downloaded the 64-bit version in your post (both to make sure they weren't just switched), and the install program is telling me I'm about ready to install the 32-bit version.  Is this just a minor error and I really am installing the 64-bit version, or are they both the same 32-bit version?

EDIT: No, the prolasso64.zip in your post above is in fact the 32-bit version.  :(

TfH

Quote from: wineaux on October 19, 2010, 10:24:03 PM
EDIT: No, the prolasso64.zip in your post above is in fact the 32-bit version.  :(

Yeah, i noticed that too when Jeremy posted this: http://processlasso.blogspot.com/2010/10/great-new-progress-and-new-daily-builds.html
I was thinking that i accidentally download 32-bit version and then my quality time with internet ended and i had to rush working. I completely forgot that and now i  re-downloaded them both, and yes. Same file with different filename.

prolasso.zip
Adler32: C5DAA938
CRC32: DF313A19
HAVAL: 68C6AE36BBDE8DA2A8D5F10C4E46B9D911A0B29B9158487869D86C081AAA7031
MD2: 0EDEF4A33F2E205688B8884FD39F7FE6
MD4: 9627AD258F398C98FBA65C5D0DDB431A
MD5: 25567E00223F74D709814AE3E8A0123A
RIPEMD-128: F6F52C705C59458E170BC0108D3B5E0D
RIPEMD-256: CB4B24FF4253C295777D4D82E3FB7B76C282F3B4914C90A0EBBFDAC40530E464
RIPEMD-320: CBD60B674E68EC69FB6AC0F5739BE7A6BB518165C0351AC335E28BF1024DD7D076B2FF318E9BF99E
SHA-1: 8A86DEBCAAE476B33206A495C4240142B0B6759D
SHA-256: 698333DDE157336B171C499880E2D01916F1A39BD20D077A2788236F391145D9
SHA-384: A6EA20E8D74CD73B3DE26E271E4618A415E2C4CF8F6A2F72FDEA00DFCC7EFF4DF2D39EEF6F1BEC6894814E38E9B3A10F
SHA-512: 6181A049919BF65C11273C7214E0175024B90E1F586C6EE28299A2D19D6E9A0DFA27A73D2330004C5B04EE7D64C08A2CA26C87BAC1FB4EF76072C0CE8D4288EB
Tiger: 380AFA9EA91DDEAF1D38FCEEE5977EE223B84602B346AF1F
Whirlpool: 80BC814F97E304A41BCCCA1EDD435DC69AAB068D4812EC30DD868FFC71C9F93D687E95F5B5BDDC217BFA122509D49D775AAEF4FFC9AB04A0ACAB742619F0F146


prolasso64.zip
Adler32: C5DAA938
CRC32: DF313A19
HAVAL: 68C6AE36BBDE8DA2A8D5F10C4E46B9D911A0B29B9158487869D86C081AAA7031
MD2: 0EDEF4A33F2E205688B8884FD39F7FE6
MD4: 9627AD258F398C98FBA65C5D0DDB431A
MD5: 25567E00223F74D709814AE3E8A0123A
RIPEMD-128: F6F52C705C59458E170BC0108D3B5E0D
RIPEMD-256: CB4B24FF4253C295777D4D82E3FB7B76C282F3B4914C90A0EBBFDAC40530E464
RIPEMD-320: CBD60B674E68EC69FB6AC0F5739BE7A6BB518165C0351AC335E28BF1024DD7D076B2FF318E9BF99E
SHA-1: 8A86DEBCAAE476B33206A495C4240142B0B6759D
SHA-256: 698333DDE157336B171C499880E2D01916F1A39BD20D077A2788236F391145D9
SHA-384: A6EA20E8D74CD73B3DE26E271E4618A415E2C4CF8F6A2F72FDEA00DFCC7EFF4DF2D39EEF6F1BEC6894814E38E9B3A10F
SHA-512: 6181A049919BF65C11273C7214E0175024B90E1F586C6EE28299A2D19D6E9A0DFA27A73D2330004C5B04EE7D64C08A2CA26C87BAC1FB4EF76072C0CE8D4288EB
Tiger: 380AFA9EA91DDEAF1D38FCEEE5977EE223B84602B346AF1F
Whirlpool: 80BC814F97E304A41BCCCA1EDD435DC69AAB068D4812EC30DD868FFC71C9F93D687E95F5B5BDDC217BFA122509D49D775AAEF4FFC9AB04A0ACAB742619F0F146

Jeremy Collake

OOps, typo in a batch file .. what can I say? Fixing now. Thanks for letting me know.

I have since republished v3.99.39 beta as THAT build, so you can also just redownload v3.99.39 beta.
Software Engineer. Bitsum LLC.

wineaux

Same problem as before with the new build.  As Dr. McCoy said many times, "It's dead, Jim!"

Scott


Jeremy Collake

Thank you for testing. That is an interesting and unexpected result. Let me rexamine this from a fresh perspective. I mean, there are only so many things that changed. I am shocked that didn't fix it. Hmm... Yea, I'll provide a debug build if it comes to that. I've got a new useful debugging procedure/model, so I can make good use of crash reports now (one of the things I've been 'wasting' time on if you were to ask a commercial proifteer). However, things are in flux right now... so we'll do that if I don't get it next build.
Software Engineer. Bitsum LLC.

Jeremy Collake

And why is my moderator Insane? ;p LOL  (referring to profile pic if it changes later when someone is reading this)
Software Engineer. Bitsum LLC.

wineaux

Quote from: jeremy.collake on October 20, 2010, 11:51:25 PM
And why is my moderator Insane? ;p LOL  (referring to profile pic if it changes later when someone is reading this)
Because he hasn't found his happy place like you have in your profile pic?

Jeremy Collake

#21
Quote from: wineaux on October 21, 2010, 10:55:13 AM
Quote from: jeremy.collake on October 20, 2010, 11:51:25 PM
And why is my moderator Insane? ;p LOL  (referring to profile pic if it changes later when someone is reading this)
Because he hasn't found his happy place like you have in your profile pic?

You know that profile pic of mine was taken on my honeymoon. It is the only time I've ever been to the Caribbean. I loved it. It was St. Croix. I like the storm on the horizon. We went during off season, so had the resort almost all to ourselves. It was wonderful. Oh, the good memories ;). Maybe someday I can afford a trip like that (last one paid for by family).

So.... I don't suppose v3.99.41 magically fixed the issue, eh? I am really perplexed by this. Can I get a copy of your INI file if this did not fix the issue? I am making RAPID progress and am so close to releasing RC1.... Well, other than this mysterious crash that appeared. It *might* have something to do with the W2K compatibility being restored, as I had to go back and use the older v9 CRT (common runtime library). Doing this with VS2010, which should be using v10 CRT might have caused some sort of complication. Dunno. I'm hoping I got the crash in .41 though...

Thanks ;)

Software Engineer. Bitsum LLC.

wineaux

Sadly, no... 3.99.41 doesn't work either.

On a wholly unrelated note:  That cross-eyed tiger has a green tree frog on its nose!

Jeremy Collake

v4.00.01 RC1 is out now. I found a serious memory leak in v3.99.38-v3.99.41 that may have contributed to the problem, I don't know. I'll continue looking for the cause and reviewing the code. I know I'm reverting some optimizations in the governor already.
Software Engineer. Bitsum LLC.

wineaux

RC1 is crashing as well.  Did an uninstall and reinstalled RC1, but it didn't help.

Jeremy Collake

Quote from: wineaux on October 22, 2010, 09:21:02 PM
RC1 is crashing as well.  Did an uninstall and reinstalled RC1, but it didn't help.

K thanks... I will figure it out. It seems to have started happening about the time I switched CRT versions, so I speculate that is related somehow. I do not know yet ;o. That is ok, I am going to hold v4 up a bit longer anyway it looks like.
Software Engineer. Bitsum LLC.

Jeremy Collake

I think I know the cause, potentially. I ran into problems during testing in XP. It came down to timing and how fast the Terminal Services process started (or didn't). The same is likely true in your case. I bet if you waited at the login prompt a minute, then logged in --- the crash might not appear. OR it could be something else.

Honestly, I'm frustrated by results of the testing of this new version. I rushed it, added too much too quickly... I don't know. It is good, solid code.. but just not fully cooked yet.
Software Engineer. Bitsum LLC.

wineaux

Trust me, I know.  My folks have a very small company that creates math educational software, and we had a devil of a time working out all the kinks on our math facts shooter game.  Especially Windows 7 64-Bit installs.  Oi vey!  Don't get me started on that one...  I (marketing sales type of guy) finally sussed out why the damned install didn't work on 64-Bit Win7.  That was months of utter hell trying to track that bug down.   :o  So I feel your pain.  I really do.

Jeremy Collake

I'm glad you can identify. I'm hoping to run into the crash myself in the hard-core testing Process Lasso will be put through in the coming days/weeks. It is much easier if I'm able to reproduce it. I have virtual machines for every Windows OS Process Lasso runs on, so test on them all -- and with enough code review really hope I see what's going on.

I wonder if there is any common third-party software on systems affected by this crash. It may security software or something Process Lasso is having a hard time with, and somehow crashing. I do not know -- yet.

Otherwise, I've got the code pretty good now. This is the one large bug that is still outstanding. A few minor issues exist, but nothing of the same caliber.
Software Engineer. Bitsum LLC.

wineaux

I'm using Vipre for security software if that helps.

Jeremy Collake

I will evaluate Vipre. I have issued a new build v4.00.07 beta that does fix an anomalous race condition at startup. I can not say whether it fixes this bug though (probably not ;o). I will continue to work and track this down. If two forum users see it, then it is more common than I'd like to think ;o.
Software Engineer. Bitsum LLC.

wineaux

OK, it crashed on start up, but it crashed a little differently.  This time the Process Lasso Management Console crashed and it was instantaneous.  In other words it crashed seconds faster than it had before.  This might give you a clue as to where the crash bug is located.

Jeremy Collake

#32
Thanks, that is a good clue. I think I know what may be going on now, but need to review the code and do some more investigation to be sure. Hopefully a final fix will come soon.

If the next build doesn't work, I will have to send a debug build. Then report to me the crash address. From that I should, hopefully, be able to infer the cause. Your LOG and INI files may also be helpful in the analysis.

Thanks again for reporting this and your continued patience.
Software Engineer. Bitsum LLC.

Hotrod

I have been using version 3.99.37b with a family subscription. I just upgraded several of my PC's to AVG 2011 Free Edition. After the initial release failed and an uninstall and reinstall and update to get past a failed database update, I found that they have added an automatic priority adjustment feature which changes AVG's priority based on user activity on the system. Also they have added a "gaming mode" boost for full screen apps. Upon running the scanner i got an exception report from PL asking if it could restart. Could something like this be conflicting with PL and causing these crashes???

Jeremy Collake

Quote from: Hotrod on October 28, 2010, 12:36:31 AM
I have been using version 3.99.37b with a family subscription. I just upgraded several of my PC's to AVG 2011 Free Edition. After the initial release failed and an uninstall and reinstall and update to get past a failed database update, I found that they have added an automatic priority adjustment feature which changes AVG's priority based on user activity on the system. Also they have added a "gaming mode" boost for full screen apps. Upon running the scanner i got an exception report from PL asking if it could restart. Could something like this be conflicting with PL and causing these crashes???

First, it is nice to see the 'core principle' of Process Lasso validated - though of course it is best if applications would manage their own priorities better. I recently wrote in a blog post that this is a trend, perhaps induced partly by Process Lasso ;). I don't know if I deleted that post or not (I get self-conscious and delete about half of them).

I do not believe an interoperability issue with their self-priority changes exists, but an interoperability issue with their protected processes could exist. I'll have to evaluate AVG too and take a look at what is going on. It is entirely possible self-protection technology on these processes is confusing the newer betas. I will check it out.

Thanks ;)
Software Engineer. Bitsum LLC.

Jeremy Collake

I've done some testing with AVG Internet Security 2011. I haven't seen any problems.

I believe this issue is due to a mistake somewhere in a synchronization object (or perhaps a lack of one). Version 4 is more multi-threaded than ever, so needs more synchronization objects than ever. I am going to continue with code review and hope I see the mistake. I'll also inject new safety mechanisms wherever the possibility of a problem might exist.

These kind of things are difficult to debug without easy reproduction. I'm surprised I haven't seen it on any my physical or virtual test beds, which only annoys me further ;o. I know, based on user reports, v4 is not ready for release yet. I'll keep plugging away at it and we'll hope to have a final soon.
Software Engineer. Bitsum LLC.

Jeremy Collake

#36
BTW, AVG's new 'Gaming mode' only applies to the Firewall. It is likely to help prioritize UDP packets and/or adjust firewall rules for games. It is nothing like Process Lasso's 'Gaming mode', something I wanted to clarify. I still applaud them lowering the priority class of their own scanner at times.

I have made a couple silent updates to v4.00.07 beta RC1. I don't think this issue is fixed, but if I got lucky maybe it is. Anyone is welcome to re-download and try it. I am *sure* I will find the cause in my code review. I don't have many talents, but I can see through code and should be able to identify the problem once I find the problematic code. This sort of 'full change review' is needed anyway. That is why I've held off on debug builds so far. If I do fail, then I will issue a debug build, as the problem may be at a deeper level than I thought (e.g. maybe switching back to the older CRT v9 had some unintended consequences).



Software Engineer. Bitsum LLC.

wineaux

4.0.08 is still crashing at the console.

Jeremy Collake

#38
Ok, thanks. I've also noticed a problem where the system tray icon is lost at startup (still) as well. I have some more work to do on the whole startup sequence, and its timing.
Software Engineer. Bitsum LLC.

wineaux

Haha!  I have found the culprit!  I have a program called DVDFABPASSKEY that loads on start-up as well.  I have had it running since before I started using Process Lasso, and the two have been coexisting well.  They shared their toys and didn't fight.  Something changed in v3.99.38 that caused ProLasso to start crashing at startup, as DVDFABPASSKEY was loading first, and the ProLasso would try to load and then crash.  I had DVDFABPASSKEY crash on me this morning, and so for giggles I decided to try and run ProLasso.  Guess what?  It ran!  I played around stopping and starting each one, and did a couple of reboots, including a final one with ProLasso running and DVDFABPASSKEY not running.  Now it seems that ProLasso is loading first at boot-up, followed by DVDFABPASSKEY, and guess what?  They are both coexisting with one another again, playing nice on the swing set, and sharing their toys.  Hopefully this will help you figure out what all is going on.  I'll go out on a limb and say that the issues that other people are having with ProLasso crashing on start-up are in fact boot load order issues.

Here's the URL for the DVDFABPASSKEY website: http://www.dvdfab.com/passkey.htm

Hopefully that may help you figure out what and where the issues lie.  I'm just happy to be running ProLasso again.  :D

Jeremy Collake

Nice work ;) Thank you for this. It actually coincides with my latest theory about what may be happening. I believe one of three things is happening:


  • Process Lasso tries to access a process not completely loaded yet, causing a crash
  • Services Process Lasso depends on haven't been fully started yet (specifically Terminal Services)
  • A process restraint event (or other type of event) is occurring during startup at just the wrong time

Unless Process Lasso was taking action on that application's process, then we'll have to assume it is #1 or #2. It makes since for this to happen during startup when several processes are being started at once, causing many to be in a half-loaded state for some time. What I'll do is add more safeguards against this.

Thanks ;).
Software Engineer. Bitsum LLC.

wineaux

So, I had ProLasso and DVDPASSKEY playing well with each other, and even had a ProLasso update that installed perfectly without having to shut down DVDPASSKEY, and both programs continued running together well.  Latest ProLasso update today broke the Console Manager again.  Shut down DVDPASSKEY and attempted to load ProLasso again.  Console Manager still crashing on startup.  Obviously it's not playing well with another program that is running.

Can't get it running at all now.  Rebooting to see what happens then.


Jeremy Collake

#42
Question: There is a Windows registry value that sets the critical section timeout value. It is at:

HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Session Manager\CriticalSectionTimeout

The default value for most (if not all) Windows NT home variants is 0x278d00 (that's hexadecimal for any layman readers). Have you run any utilities that may have tweaked that value? If that value is too low, then a critical section can time out, raising an exception. I changed my own critical section settings last build, and that change could cause some of them to end up waiting longer (but saving CPU cycles). Please check this registry value if you get a chance. Its probably not it, but I just want to be sure. Some of these optimization utilities do things they think will help, but actually have consequences. That is one value that should NOT EVER be decreased.

May or may not be the cause here, but is a note for other readers even if not applicable to you.

Software Engineer. Bitsum LLC.