Author Topic: XWorkplace and missing folder contents...  (Read 1646 times)

Dariusz Piatkowski

  • Sr. Member
  • ****
  • Thank You
  • -Given: 12
  • -Receive: 7
  • Posts: 302
    • View Profile
XWorkplace and missing folder contents...
« on: October 01, 2017, 06:05:35 pm »
So I've been struggling with this for a while now. Running XWP 1.0.11 here, SMP system with ACPI 3.23.06 release and 5 of the 6 CPU cores enabled. The 6th is off-line since attempting to bring WPS up with it enabled results in the dreaded WPS blue-screen hang.

Alright, so the issue is that any folder views, but most evident with the Drives/filesystem objects, stop populating. Initially, they will show objects, but eventually as I open new ones they halt, so the folder remains blank. As soon as this happens, even though it may have happened with a filesystem folder first the remaining WPS folder objects stop populating as well. The result is that my OS/2 box simply becomes unusable, I can no longer access any of my data or program objects. Re-boot is the only solution.

I have not found the root cause yet. I tried many things, playing with various setups, Lazy icons on/off, etc, etc. What I did notice is that putting my system into a single-core mode (by shutting off the remaining 4 cores) appears to work fine, so that implies the XWP folder functionality somehow runs into an issue in SMP environments.

There is a LOG entry generated in the xwptrap.log file (attached to the post, it's a bit lengthy), but that log does not tell me much.

Since 1.0.11 upgrade I have attempted to use the XView option...maybe that is causing the problem? Hard to tell, because even when I do not open any folders which use this mode the issue does show up anyways.

Most recently I found the following issue in the 'Workplace Shell' object when reviewing the 'Sort' tab I find multiple occurencies of the 'Real Name' option, see screenshot in the attachment. I took some time to go through the INI files but could not find where XWP gets this info. The HELP system gives the following info:

Sort by real name: this sorts a folder alphabetically according to file-system names. Since only file-system objects have real names, abstract objects will not be sorted properly.

...needless to say, I have not been able to figure out why I see multiple entries. This may be a whild goose chase, but I have nothing else to go on at the moment.

So the question to everyone: have you seen this behaviour, can you shed any light on what else I should look at?

Thanks!



Andreas Schnellbacher

  • Sr. Member
  • ****
  • Thank You
  • -Given: 1
  • -Receive: 15
  • Posts: 257
    • View Profile
Re: XWorkplace and missing folder contents...
« Reply #1 on: October 01, 2017, 06:12:51 pm »
Have you activated "Turbo Folders"?

Dariusz Piatkowski

  • Sr. Member
  • ****
  • Thank You
  • -Given: 12
  • -Receive: 7
  • Posts: 302
    • View Profile
Re: XWorkplace and missing folder contents...
« Reply #2 on: October 01, 2017, 06:19:55 pm »
Hi Andreas,

Have you activated "Turbo Folders"?

At some point in time I did try that, but I wasn't too crazy about how the filters & WPS associations setups were done so I ended up shutting it off after a trial period.

Do you think I should pursue this again? It's been a while, I know it was one of my debugging efforts in hopes of finding the 'cure' to my problem...as best as I can remember the blank folder issue continued there as well.

ivan

  • Hero Member
  • *****
  • Thank You
  • -Given: 0
  • -Receive: 24
  • Posts: 657
    • View Profile
Re: XWorkplace and missing folder contents...
« Reply #3 on: October 01, 2017, 10:27:54 pm »
I wonder if 'sort by real name' and 'always maintain sort order' are in conflict. 

The 'always maintain sort order' is the first thing I switch off in any new install.  It started causing problems with our AMD systems way back and it still annoys me to this day.

Rich Walsh

  • Newbie
  • *
  • Thank You
  • -Given: 0
  • -Receive: 4
  • Posts: 31
    • View Profile
Re: XWorkplace and missing folder contents...
« Reply #4 on: October 01, 2017, 11:26:38 pm »
First, Dariusz, let me apologize for not responding to your email from Thursday. To answer your question, yes I have looked into this problem but I can't find anything.

Here's how I see it: a lot of people use XWP but you're the only one I recall having this particular issue. Perhaps that's because you may be the only person who has a particular, possibly unique, collection of add-ons that cause or at least contribute to the onset of this problem. Rather than attributing the cause to the lowest common denominator, XWP, why don't you look at the other stuff you have installed - particularly stuff that few people use.

For instance, the first thing I noticed when I looked at that very odd screen shot were those doo-dads on the titlebar. I've never seen them before - what puts them there? What other WPS "enhancements" are you using? One of them may be causing the problem or simply triggering a seldom-seen bug in XWP but we'll never know until you do the legwork to identify the proximate cause. Anything that registers a WPS class should be high on your list of suspects.

Start by uninstalling a few of the the least common or least useful add-ons, then wait a while (a week?). If the problem still happens, you know what isn't to blame, so you can reinstall them and try removing other stuff. Once it looks like the problem has been banished, start adding things back until it reappears. Yes, it may take a while (a month or more), and you may find yourself inconvenienced by the lack of this or that - but after a day or two of disorientation you may also find that a lot of this stuff isn't as must-have as it once seemed.

Doug Bissett

  • Hero Member
  • *****
  • Thank You
  • -Given: 1
  • -Receive: 38
  • Posts: 937
    • View Profile
Re: XWorkplace and missing folder contents...
« Reply #5 on: October 02, 2017, 12:14:20 am »
Quote
Here's how I see it: a lot of people use XWP but you're the only one I recall having this particular issue.

Bull. I gave up complaining about this issue, years ago, because I "am the only one with the problem". I have tried many things to try to isolate the problem, but it still happens, once in a while, with ArcaOS. It will be fixed, temporarily, by a WPS restart, but it always comes back in a few minutes (usually long enough to save my work, and reboot).

Reboot is the only thing that fixes it for any length of time. I don't, and never have, used any other add ons (except to try them), except for DragText. This has happened a few minutes after boot, and the system has run for a week, without seeing the problem. It has persisted over many machines, many versions of XWP, and from before I started to use eCS. In fact, it may have nothing, at all, to do with XWP, but it is the most likely suspect. I suppose I need to open a new issue about it, but I can't add too much to the story. The best that I can hope for, is that one of the main developers will see the problem, figure it out, and fix it.

Rich Walsh

  • Newbie
  • *
  • Thank You
  • -Given: 0
  • -Receive: 4
  • Posts: 31
    • View Profile
Re: XWorkplace and missing folder contents...
« Reply #6 on: October 02, 2017, 06:30:53 am »
Bull. I gave up complaining about this issue, years ago, because I "am the only one with the problem".

Thanks for advising. My apologies if I didn't recall that you bitched and moaned about this years ago - one rant tends to blend into another.

Happily, an unusual confluence of events today has opened a whole new avenue to explore - one that has nothing to do with XWP or the WPS. As you know, I got to see part of your config.sys today and the many, many filesystems you load. Perhaps that's where the problem lies. As I understand it, all filesystem requests go through the kernel's filesystem router, then out to the appropriate one. If something were to get messed up there, it's easy to see how some or all filesystem activity would grind to a halt.

Based on the excerpt of config.sys you posted, I see that you load:
  • JFS
  • HPFS
  • CDFS
  • FAT32
  • UDF
  • STUBFSD / AEFS
  • STUBFSD / ISOFS
  • NDFS
  • TONIGY
Dariusz, do you use any of the less common ones on this list?

Quote
The best that I can hope for, is that one of the main developers will see the problem, figure it out, and fix it.

That's pretty amusing...

roberto

  • Sr. Member
  • ****
  • Thank You
  • -Given: 14
  • -Receive: 6
  • Posts: 327
    • View Profile
Re: XWorkplace and missing folder contents...
« Reply #7 on: October 02, 2017, 10:09:20 am »
Hi Dariusz
I have this problem too, but in my case it is in a peer to peer network.
The server is a pc hp z400 = pc1 Arca50 with Broadcom BCM57xx(x)/BCM590x MMBGE.os2
The computer with this problem is a hp z400 = pc2 Arca50  with Broadcom BCM57xx(x)/BCM590x MMBGE.os2
The computer that I copied with the wpshell copy object (s) is pc3
The three computers have Arca5.0
With the computer pc3 I made a copy to pc1, the pc1 from the pc1 sees the folder and its contents.
The pc3 sees the folder well and its contents looking at the pc1
But pc2 sees the folder on pc1, but does not see the content.
Pc2 fails with wpshell, fails with lcm, and fails with fc2 2.4.

This pc2 gives me strange bugs with the dbexpert, which in some applications that uses dbexpert
proves that the dbexpert is closed without any reason, to justify it.
Saludos

Andreas Schnellbacher

  • Sr. Member
  • ****
  • Thank You
  • -Given: 1
  • -Receive: 15
  • Posts: 257
    • View Profile
Re: XWorkplace and missing folder contents...
« Reply #8 on: October 02, 2017, 11:40:58 am »
when I looked at that very odd screen shot were those doo-dads on the titlebar. I've never seen them before - what puts them there? What other WPS "enhancements" are you using?
BTW: I remember that X-it adds such titlebar buttons.
« Last Edit: October 02, 2017, 11:55:05 am by Andreas Schnellbacher »

Lars

  • Full Member
  • ***
  • Thank You
  • -Given: 0
  • -Receive: 21
  • Posts: 162
    • View Profile
Re: XWorkplace and missing folder contents...
« Reply #9 on: October 02, 2017, 03:31:30 pm »
Bull. I gave up complaining about this issue, years ago, because I "am the only one with the problem".

Thanks for advising. My apologies if I didn't recall that you bitched and moaned about this years ago - one rant tends to blend into another.

Happily, an unusual confluence of events today has opened a whole new avenue to explore - one that has nothing to do with XWP or the WPS. As you know, I got to see part of your config.sys today and the many, many filesystems you load. Perhaps that's where the problem lies. As I understand it, all filesystem requests go through the kernel's filesystem router, then out to the appropriate one. If something were to get messed up there, it's easy to see how some or all filesystem activity would grind to a halt.

Based on the excerpt of config.sys you posted, I see that you load:
  • JFS
  • HPFS
  • CDFS
  • FAT32
  • UDF
  • STUBFSD / AEFS
  • STUBFSD / ISOFS
  • NDFS
  • TONIGY
Dariusz, do you use any of the less common ones on this list?

Quote
The best that I can hope for, is that one of the main developers will see the problem, figure it out, and fix it.

That's pretty amusing...

Every volume ("drive letter") is presented to every IFS on the "FS_MOUNT" call to the IFS. Once an IFS claims responsibility for a volume ("drive letter") it does so for good (until next reboot, of course). From then on, the kernel knows what file system Driver (IFS) to invoke for any directory or file related access.

What is important to know is that the FAT32.IFS Version that delivered with AN was badly broken and far from being ready. It would be a good idea to pull the latest release from Netlabs.
Nonetheless, assuming that FS_MOUNT of FAT32.IFS is not doing something fundamentally wrong, only FAT32 partitions (and FAT12/FAT16 and exFAT if you enabled that in FAT32.IFS) should suffer from file system driver errors.

Dariusz Piatkowski

  • Sr. Member
  • ****
  • Thank You
  • -Given: 12
  • -Receive: 7
  • Posts: 302
    • View Profile
Re: XWorkplace and missing folder contents...
« Reply #10 on: October 02, 2017, 05:31:28 pm »
when I looked at that very odd screen shot were those doo-dads on the titlebar. I've never seen them before - what puts them there? What other WPS "enhancements" are you using?
BTW: I remember that X-it adds such titlebar buttons.

That's correct, Andreas is right on the money here. There are literally just a handful of X-it things I'm using now though:
1) the icon showing 3 windows is just a short-cut to TaskList, and that is replaced with the superhandy LSwitcher TaskBar
2) the two triangles/arrows gives you the window roll-up feature (pretty handy when the Z-order changes and you simply want to get back to the background stuff), but alas, I am not aware of anything else that implements this

Anyways, for a long time I was hoping both of these would be replaced with XWP features, but as Rich's comment below correctly points out, we just do not have the massive developer base to make these things happen.

So...in my further debugging of this issue I think I stumbled across something that may be key to this...I will do a separate post though to keep that focused on that single finding.

Dariusz Piatkowski

  • Sr. Member
  • ****
  • Thank You
  • -Given: 12
  • -Receive: 7
  • Posts: 302
    • View Profile
Re: XWorkplace and missing folder contents...
« Reply #11 on: October 02, 2017, 05:42:19 pm »
Hi Rich,

First, Dariusz, let me apologize for not responding to your email from Thursday. To answer your question, yes I have looked into this problem but I can't find anything.

Here's how I see it: a lot of people use XWP but you're the only one I recall having this particular issue. Perhaps that's because you may be the only person who has a particular, possibly unique, collection of add-ons that cause or at least contribute to the onset of this problem. Rather than attributing the cause to the lowest common denominator, XWP, why don't you look at the other stuff you have installed - particularly stuff that few people use...

No worries regarding the delays, believe me, I understand people have real lives outside of this great hobby of ours. If anything though, I would have appreciated just seeing that very straight response back in late June when I sent you my last email with the trap logs, otherwise it's a bit of a vacuum thing and it's easy to assume the last person one has delt with on a particular issue has moved on to "greener" pastures...LOL!

So yes, you make an excellent point. This is actually something I had gone through as well. I did spend massive amounts of time trying to find the root cause here. That meant I removed all the various add-ons until I literally stripped my WPS down to just "out of the box" type install, the key difference being XWP. The issue was still there though...but recently, it seems like it has gotten much worse though.

That is ultimately what prompted me to look for feedback from the OS/2 Forum community...as you stated, it may be something specific to maybe core OS/2 components in my install even?

Dariusz Piatkowski

  • Sr. Member
  • ****
  • Thank You
  • -Given: 12
  • -Receive: 7
  • Posts: 302
    • View Profile
Re: XWorkplace and missing folder contents...
« Reply #12 on: October 02, 2017, 06:15:15 pm »
OK, so now the possible root cause description...this is based on the last change I made and the DAY/NIGHT type of a difference I am seeing here.

If we go back a few months ago the FF 45.x release was starting to gain a bit of movement, to the point where I deemed it stable enough to move from 38.x to the test 45.x builds. One thing that has plagued us in that 45.x release is the CPU spikes issue, or specifically how after extended use (about 5 hrs here) the system just appears to literally slow down to a crawl to a point where only a FF process kill and re-start will get things back up and moving.

OK, so why does this matter?

Well, back then as my 5 cores would get 100% pegged and system became unresponsive I thought I should follow the recommendation of the XWP Help book and adjust the 'Worker thread' setting to the next higher value from the default, which is "idle +31". This means that I set this to 'regular'.

The XWP Help specifically states:

"...For most purposes, you should leave this setting to the default value, which is "idle +31". This will allow the Worker thread to perform its tasks in the background without a noticeable slowdown of the user interface. However, if you frequently run processes which hog up your CPU (that is, if your CPU load is always 100% because of, e.g., certain Windows applications running), you might want to raise this to "regular"..."

Now, as I was reviewing this and watching the various thread priorities I started questioning whether something was causing a race-type condition...or perhaps a much simpler priority driven "incompatibility". This idea popped into my head as I was reviewing the OS/2 Toolkit documentation and specifically looking at the MARKEXE documentation and how code design decisions impact execution in a SMP enviornments (remember, I previously noticed that disabling the 4 of y 5 cores seemed to produce a working environment where the problem was not occuring).

Here is the specific section of the MARKEXE information:

"...The following compatibility's requirements should be considered before running an application in multi-processor mode:

An application or associated subsystem must not use the 'INC' instruction as a semaphore without prepending a 'LOCK' prefix. On a UniProcessor (UP) system this instruction can be used as a high-performance semaphore without calling any other operating system service if the semaphore is free and when the semaphore is clear and there are no waiters for the semaphore. Because the INC instruction cannot be interrupted once started, and because the results would be stored in the flags register which are per thread, then it could be used safely as a semaphore.

In an OS/2 for SMP environment, this technique will not work because it is possible that two or more threads could be executing the same INC instruction, receiving the same results in each processor's or thread's flag register and thinking that they have the semaphore..."


Alright, to make it clear, I did not use MARKEXE to flag XWP as SMP un-safe, I am merely showing the above as a point of information.

So with this XWP Thread Worker change made, and getting me back to the XWP default priority setting what am I seeing? Well, I have several folders open, System Drives object, single Drive object and so on. None of them (so far, knock on wood...LOL) are running into this problem, and for me that is absolutely GREAT news!!!

I am not saying this has gotten me a working sytem, but in previous instances the folders would typically  stop working in a manner of 5-10 mins...in other situations, they would be OK for several hours and than all of a sudden - BRICK WALL!

My machine has been up and running, with some pretty heavy FS use for a tad over 24 hrs...I will provide further updates as some additional time lapses and I have had the opportunity to re-boot and try several iterations. I've attached a screenshot of the 'XWorkplace Setup' object showing the Threads tab. Notice that the Worker thread is lower priority than SplitPopulate, previously though it would have been the same priority.

For now though, Rich, does this help out in any way in terms of what may be going wrong if one forces the higher priority?

Dariusz Piatkowski

  • Sr. Member
  • ****
  • Thank You
  • -Given: 12
  • -Receive: 7
  • Posts: 302
    • View Profile
Re: XWorkplace and missing folder contents...
« Reply #13 on: October 03, 2017, 08:11:23 pm »
...So with this XWP Thread Worker change made, and getting me back to the XWP default priority setting what am I seeing? Well, I have several folders open, System Drives object, single Drive object and so on. None of them (so far, knock on wood...LOL) are running into this problem, and for me that is absolutely GREAT news!!!

I am not saying this has gotten me a working sytem, but in previous instances the folders would typically  stop working in a manner of 5-10 mins...in other situations, they would be OK for several hours and than all of a sudden - BRICK WALL!...

Well, unfortunately this only turned out to be a "short but sweet" victory!

Alas, upon a re-boot I am back to where I started...same issue as before.

I suspected that maybe something about the shut-down/re-boot cycle is causing a problem, so I reverted back to my OS2SYS.INI & OS2.INI files (I regularly back them up during re-boots) to get me back to the starting point where I saw this issue go-away. No luck though, even with that re-set the issue was immediately apparent.

Sooo...theorizing about SMP thread contention aside, there must be something else at play causing the problem.

Rich Walsh

  • Newbie
  • *
  • Thank You
  • -Given: 0
  • -Receive: 4
  • Posts: 31
    • View Profile
Re: XWorkplace and missing folder contents...
« Reply #14 on: October 05, 2017, 10:21:26 am »
I'm sorry to see that this problem still persists. I waited to reply until I had a chance to look at that worker-thread code.

Like much of XWP, It uses a mutex semaphore to protect some data that other threads modify. While this is perfectly OK, it's willing to wait indefinitely for its sem to become available. While I don't think the problem lies on this thread, it's easy to see how these indefinite waits could lead to a deadlock. Both threads A and B need to own semaphores X and Y to proceed. A holds one while B holds the other, so both wait forever for a semaphore the other will never release. If one of those threads handles populating folders, the WPS will grind to a halt. Where this is happening - if that's actually what's happening - remains a mystery.

BTW1... Have you ever tried disabling XWP to confirm that it really is the problem? It's easy: unlock 'xfldr.dll', rename it, then restart the Desktop. Don't rely on XWP to disable itself - this is the only method short of uninstalling that is guaranteed to do so.

BTW2... I'd still like to know what filesystems you load...