• Welcome to OS2World OLD-STATIC-BACKUP Forum.
 

News:

This is an old OS2World backup forum for reference only. IT IS READ ONLY!!!

If you need help with OS/2 - eComStation visit http://www.os2world.com/forum

Main Menu

OS/2 Single Input Queue

Started by PAUL555, 2008.01.19, 12:54:10

Previous topic - Next topic

PAUL555


  Hi All , 

     Has  the  single  input  queue  problem  been  completly  resolved  in  Ecomstation  ?  I  know  that  IBM  attempted  to  fix  this  with  Fixpack  17  for  OS/2  Warp  3.0 . Was  it  ever  resolved  or  is  there  just  a  workaround  ? 

Regards

Nirmal

Andi710

AFAIK, nothing substantial has been changed in this respect.

mobybrick

I have to say that errant apps hogging the entire OS has been much less of a problem since Warp 4 - and eComStation further enhances this with the integration of CADH.

The ability to 'kill' errant processes allows queue hogs to be (usually) removed.

Regards,
Moby.

Saijin_Naib

Many apps still simply refuse to end, either with TOP or Window list and using hard kill, necessitating a reboot to get the WPS back.

mobybrick

Curious. I normally find that with a 30-second wait (for the SIQ 'fix' to fully complete) most apps close. The thing to do is to wait for the 30-second timeout upon hang, before trying to kill the application.

Regards,
Moby.

Pete

Quote from: PAUL555 on 2008.01.19, 12:54:10

  Hi All , 

     Has  the  single  input  queue  problem  been  completly  resolved  in  Ecomstation  ?  I  know  that  IBM  attempted  to  fix  this  with  Fixpack  17  for  OS/2  Warp  3.0 . Was  it  ever  resolved  or  is  there  just  a  workaround  ? 

Regards

Nirmal


I'm not quite sure if it ever got classified as "Resolved" but it certainly does not seem to cause the problems that it did with W3 and early W4 installations.

Might be worth using Google to find a few posts on the topic.

Regards

Pete

Criguada

The SIQ problem has not been solved, but a thorough workaround ha been put in place. The only way to solve the SIQ problem would be to implement a MIQ (multiple input queue); this has been judged as infeasible by IBM because it would break a lot of the existing applications.
The SIQ fix has been doing a good job, and eCS has very few problems in my expercience due to the SIQ. Those who think they are experiencing the SIQ problem, mostly are experiencing other bugs (in the WPS or in other apps/classes).
One example is the problem of unkillable tasks: this is a completely different problem/limitation in OS/2, which can not easily be fixed without rewriting the kernel. I can offer a technical explanation for the curious.
When a task hangs, and it becomes unkillable due to this problem, it often triggers the problem in the SIQ too. But since you're not able to kill the task, you're not able to free the queue as well. At this point, you'll see OS/2 responding with a window that says more or less this "process xxx has stopped responding, do you want to end it?". This is the SIQ fix kicking in. But if you say "kill it!", OS/2 won't because it is not able to. So you're stuck in a circle, where OS/2 continually asks you if you want to kill the task, you say "yes", and nothing happens. Only way out: reboot.
This does not happen frequently, especially if you try not to use experimental software. If you have software that frequently hangs this way (i.e. it becomes unkillable) a solution sometimes is to immediately kill it with a "kill -9", without trying first with a normal kill.

Hope this helps
Bye
Cris

RobertM

#7
My only remaining problem is with Firefox - usually during shutdown (and usually if gMail is open on a tab). Firefox hangs (someplace) during shutdown, and nothing responds on the machine (WPS, xWP, etc)... a few CTRL-ESC or ALT-ESC makes the SIQ workaround notice and take action (in 5-10 seconds - but I have the "Focus Change Sensitivity" set to 5 or less - not the default 20)... once the SIQ worker notices, it will prompt me with "thread 192 (or whichever) has stopped responding, do you want to end it?" - at which time I get the "OK" and often no cancel choice. Selecting "OK" kills it, and the remnants of Firefox, (and usually frees a ton of memory - often more than is normally deallocated when Firefox shuts down normally) and all is well.

Occassionally, nothing seems to kill Firefox when it hangs... but for me, that is rare... in the instances where Firefox does hang, the kill/unkillable (requires reboot) ratio is probably 10:1.

On the rare instances Firefox is not killable, I have found that I *need* to hit C-A-D (twice1 because I have cadh installed - and it doesnt come up at all even tho the first CAD thinks it switched to it) to reboot and need to do it before I keep clicking on other things (either with the keyboard or mouse) or the system will not respond to the C-A-D keys and just beep when they are pressed, forcing a hard-power-off situation (or CTRL-ALT-NUMLOCK-NUMLOCK - which has always worked for me in every lockup situation - and is easier than reaching for the power switch...)




As an interesting side note, if you do get the "Process/Thread" (whatever) has stopped responding..." message and no cancel button - but for whatever reason want to hit cancel (ie: it is a resouce/GUI hungry process that you dont want to kill, or the SIQ worker has identified the wrong app, etc), then (assuming that window has focus) simply hit the escape key. Doing so seems to do the same thing as hitting the cancel button (when it is on the dialog box).

I am kinda curious why sometimes there is - and sometimes there isn't - a cancel button... but since the escape key seems to work the same, it's not really a big deal to me...




-Rob

Note 1: When trying to bypass CADH by double pressing CTRL-ALT-DEL, I find I need to hold CTRL-ALT, then press DEL, then release ALL of them and repeat to actually reboot. Oddly, holding CTRL-ALT, and hitting DEL twice (without letting go of CRTL and ALT) will not reboot any of my systems...



|
|
Kirk's 5 Year Mission Continues at:
Star Trek New Voyages
|
|


RobertM

Oh... there is one other thing... the HTCIGBIN.EXE that Lotus Domino Go Webserver runs to handle CGI processing is unkillable by any method I have tried (though I need to check and ensure that kill-9 support is on that machine). At least one is always unkillable by my human intervention - even after shutting down the web server - but oddly, if I want to wait for the server timeout, the server can almost always kill the hung cgi process.


|
|
Kirk's 5 Year Mission Continues at:
Star Trek New Voyages
|
|


Robert Deed

When firefox hangs majorly solid and requires the double CTRL-ALT-DEL it has nothing to do with the Input Queue..  it has to do with Innotek's font engine running out of system resources (basically just killing the whole graphics subsystem ).  Obviously this is why IBM never was big on anyone hooking the graphics subsystem in this way.  I've noticed that this happens more or less after long periods of uptime especially if I am reading and writing in alot of large forums (some coding forums I use, along with technical assistance sites I participate in).  I'm not saying you thought that this had anything to do with the SIQ but I can say I am 99% sure of this, since when I removed the font engine it stopped occuring.  Of course I reinstalled it because I would take a crash every month or two over unreadable small fonts on my 20" 1600x1200 display =)  (I'm sure my large display size also increases the frequency of this problem).

Quote from: RobertM on 2008.03.20, 06:27:50
My only remaining problem is with Firefox - usually during shutdown (and usually if gMail is open on a tab). Firefox hangs (someplace) during shutdown, and nothing responds on the machine (WPS, xWP, etc)... a few CTRL-ESC or ALT-ESC makes the SIQ workaround notice and take action (in 5-10 seconds - but I have the "Focus Change Sensitivity" set to 5 or less - not the default 20)... once the SIQ worker notices, it will prompt me with thread 192 (or whichever) has stopped responding, do you want to end it?" - at which time I get the "OK" and often no cancel choice. Selecting "OK" kills it, and the remnants of Firefox, (and usually frees a ton of memory - often more than is normally deallocated when Firefox shuts down normally) and all is well.

Occassionally, nothing seems to kill Firefox when it hangs... but for me, that is rare... in the instances where Firefox does hang, the kill/unkillable (requires reboot) ratio is probably 10:1.

On the rare instances Firefox is not killable, I have found that I *need* to hit C-A-D (twice1 because I have cadh installed - and it doesnt come up at all even tho the first CAD thinks it switched to it) to reboot and need to do it before I keep clicking on other things (either with the keyboard or mouse) or the system will not respond to the C-A-D keys and just beep when they are pressed, forcing a hard-power-off situation (or CTRL-ALT-NUMLOCK-NUMLOCK - which has always worked for me in every lockup situation - and is easier than reaching for the power switch...)




As an interesting side note, if you do get the "Process/Thread" (whatever) has stopped responding..." message and no cancel button - but for whatever reason want to hit cancel (ie: it is a resouce/GUI hungry process that you dont want to kill, or the SIQ worker has identified the wrong app, etc), then (assuming that window has focus) simply hit the escape key. Doing so seems to do the same thing as hitting the cancel button (when it is on the dialog box).

I am kinda curious why sometimes there is - and sometimes there isn't - a cancel button... but since the escape key seems to work the same, it's not really a big deal to me...




-Rob

Note 1: When trying to bypass CADH by double pressing CTRL-ALT-DEL, I find I need to hold CTRL-ALT, then press DEL, then release ALL of them and repeat to actually reboot. Oddly, holding CTRL-ALT, and hitting DEL twice (without letting go of CRTL and ALT) will not reboot any of my systems...



RobertM

#10
I wasnt sure what the cause was - but that does seem to fit with past experiences... some "auto-killable" Firefox issues (when it crashes all by itself - but is restartable) end up having no IFE support... but at the same time, other than non-anti-aliased fonts, FF seems to run with no further crashes, and without leaking nearly as much memory.

I find running FF with the version of IFE on this machine, if I leave it up with a few tabs open (but do nothing with it), and leave for a number of hours, I am almost always guaranteed to come back and find the machine has no physical memory left - and the swap file is growing in use massively. I have not seen that issue without IFE open (I dont think I have seen it with the other version of IFE - but that one wont install on this machine - at least not in a way that leaves the machine bootable... trying to install it leaves me in a situation where I am guaranteed to never see the desktop again - until I boot to command prompt and remove the IFE files (or replace them with the old version).

-Robert


|
|
Kirk's 5 Year Mission Continues at:
Star Trek New Voyages
|
|


Robert Deed

You defiantly have some interesting hardware there.  IFE has never made my system fail to boot, though it has made it fail to run firefox before (though oddly enough other apps which used it worked fine).  I don't get any massive memory leaking if I leave firefox open, I've done so for weeks (while I was on vacation).  I would love to know why this happens on some systems and doesn't on others.  Especially since I have a big project planned and I am slowing starting to flip flop from OS/2 to XP for the platform of choice.  Most of the end user interaction will be done through browser and I really need a stable browser and while firefox/2 is more stable then any browser I have used it seems to have problems on others peoples systems (and the hardware I am using isn't easily reproducible today)   Obviously I want to come to the bottom of this since I will be providing the hardware and technical support for the project. 

Quote from: RobertM on 2008.03.20, 06:50:49
I wasnt sure what the cause was - but that does seem to fit with past experiences... some "auto-killable" Firefox issues (when it crashes all by itself - but is restartable) end up having no IFE support... but at the same time, other than non-anti-aliased fonts, FF seems to run with no further crashes, and without leaking nearly as much memory.

I find running FF with the version of IFE on this machine, if I leave it up with a few tabs open (but do nothing with it), and leave for a number of hours, I am almost always guaranteed to come back and find the machine has no physical memory left - and the swap file is growing in use massively. I have not seen that issue without IFE open (I dont think I have seen it with the other version of IFE - but that one wont install on this machine - at least not in a way that leaves the machine bootable... trying to install it leaves me in a situation where I am guaranteed to never see the desktop again - until I boot to command prompt and remove the IFE files (or replace them with the old version).

-Robert


RobertM

I'm wondering it it is gMail causing that (with it's continuous background operations) thus calling tons of redraws through IFE thus causing it to eventually crash - all the while either IFE or some part of Firefox is using more space (caching the ever changing gMail data possibly? as if I manually kept loading page after page?)

Not sure... though I do know FF3b3 handles gMail a lot nicer (and a LOT quicker).

-Rob


|
|
Kirk's 5 Year Mission Continues at:
Star Trek New Voyages
|
|


Robert Deed

Amusingly enough it was gmail that I had left opened, however I don't recieve a whole lot of mail on the account in question.  I have not yet tried firefox3  I used an earlier build but had abandoned it due to lack of plugin support.

Quote from: RobertM on 2008.03.20, 07:03:19
I'm wondering it it is gMail causing that (with it's continuous background operations) thus calling tons of redraws through IFE thus causing it to eventually crash - all the while either IFE or some part of Firefox is using more space (caching the ever changing gMail data possibly? as if I manually kept loading page after page?)

Not sure... though I do know FF3b3 handles gMail a lot nicer (and a LOT quicker).

-Rob

Saijin_Naib

One sure-fire way to lock up Firefox and subsequently OS/2 is to click the 1-click Bid link on an ebay auction page. That cost me an item once :C Oh well, thats Firefox's support, or maybe not, but now I know about it, I dont click it anymore. IFE is rather unstable, but 2.60 beta seems to be worlds better than the 2.40 final. I wish Innotek would open the IFE so Netlabs could work its magic on it and make it not garbage, and not a cause for insane ammounts of OS/2 issues, like it currently is.

FF3 is much leaner on memory usage, though any of the Firefox will leak memory if you leave them for a while. This is a documented case, and is sorta well known. Im sure any program leaks if left running long enough though.