Page 4 of 6 FirstFirst 123456 LastLast
Results 61 to 80 of 118
  1. #61  
    Hello,mobile accomplishers...

    Do we really need jstop install? I don't think it does anything for my palm pre. (In my opinion)

    Now that compcache kernel is here.... is jstop still a necessity?
    Last edited by rush; 06/17/2010 at 06:09 AM.
  2. TomD's Avatar
    Posts
    908 Posts
    Global Posts
    1,047 Global Posts
    #62  
    Guys,

    Is there a way to write a tool that can track memory leaks?

    That would be an awesome tool for both users and developers.

    How do they track memory leaks on other platforms?

    I'm hoping that the amazing brain power and innovative people here can come up with something.
    Tom
    Pilot Prō --> M100 --> Trēō 600 --> Trēō 700p -- > Prē
  3.    #63  
    UberKernel 1.4.x-40 is now in alpha testing, along with Govnah 0.4.6

    -- Rod
    WebOS Internals and Preware Founder and Developer
    You may wish to donate by Paypal to donations @ webos-internals.org if you find our work useful.
    All donations go back into development.
    www.webos-internals.org twitter.com/webosinternals facebook.com/webosinternals
  4. #64  
    Quote Originally Posted by rwhitby View Post
    UberKernel 1.4.x-40 is now in alpha testing, along with Govnah 0.4.6

    -- Rod
    On my way to test it out... We appreciate the hard work and effortfulness of making this colabaration a success.
  5.    #65  
    UberKernel 1.4.x-43 (for Pre and Pixi) and Govnah 0.4.7 should be in the alpha testing feeds soon (as soon as the autobuilder finishes building them), along with PsychoKernels F102A and F104A and SbromwichKernel.

    -- Rod
    WebOS Internals and Preware Founder and Developer
    You may wish to donate by Paypal to donations @ webos-internals.org if you find our work useful.
    All donations go back into development.
    www.webos-internals.org twitter.com/webosinternals facebook.com/webosinternals
  6. #66  
    Quote Originally Posted by rwhitby View Post
    UberKernel 1.4.x-43 (for Pre and Pixi) and Govnah 0.4.7 should be in the alpha testing feeds soon (as soon as the autobuilder finishes building them), along with PsychoKernels F102A and F104A and SbromwichKernel.

    -- Rod
    Rod, UnixPsycho, and everyone else..... THANKS. Finally just decided to throw the testing Kernel on my phone, only running 500/600 but the compcache does what I need it to do. Have seen some lag, but that's with a 3D game and multiple apps open, once in the game it runs as smooth as before, but no more restart just to get it to load. Between webos and my linux laptop, maybe I will finally spend some more time learning about command line now that I have a reason too.
  7. #67  
    with the old Palm OS they created an awesome way of testing programs out to find memory leak, bugs, etc. They were gremlins, and were just a great idea. It would simulate taps all over the screen and basically try out every option in a program over and over again. I wish they had them again! I use to program from WinMO and always had memory troubles, it was one reason I left WinMo.
    Bring back the gremlins, please
  8. #68  
    Quote Originally Posted by supercluver View Post
    Well I'm just glad others (especially you since you are so knowledgeable) are seeing this same situation. I'll try flushing the swap now that I have that ability with Govnah and UberKernel and see what that does for me. And I'm guessing there is probably some memory leak. Maybe TweetMe or Weather Window.... Or Palm's Facebook Beta.... guess I have investigating to do if no one knows of a memory leak with those apps....
    +1000000

    We still stand behind our findings that poor coding is the reason for TMC errors. Unfortunately the constant new profiles and $$ for apps has become prohibitive for our cause....but our findings remain consistent that on a factory fresh Pre (sprint), you can run ANY number of "Stock" apps and be able to run the 3D game of your choice, close it, run any additional number of stock apps, and run the 3D game again without issue.

    Once you begin to introduce 3rd party apps, you will most certainly begin seeing TMC's. The main issue is not "outing" those devs who just may not know, but also having technical evidence to present. (Maybe we'll just upload all our data to WebOSInternals )
  9. #69  
    you know, thinking about it.. 1 MB on Palm OS was pretty big..
    http://<font color="Navy">Poll: Requ...r webOS</font>

    Coming soon-Calm Alarm
    Quote Originally Posted by SirataXero View Post
    Palm Pimp.
    Comes with 2 free Pixis.
  10. #70  
    Holy rusted bawls, batman! As I continue debating wether or not to try out the Evo, developments like these further make me want to stay with my Pre (of which I'm on my sixth now) and wait it out till the next Palm flagship. THREE CHEERS!
  11. #71  
    What should be done is make a kill script that can kill all processes that are outside of the scope of a normal startup.

    Linux is very good at managing memory, better then Windows in every sense. If we are closing apps and they are still remaining in memory, this is an app issue.

    I could write a script, a better option would be some of our fine devs to make an app that can do this. A phone reset is nothing more then returning the device to the "start" state. This could be done with a script with no issue, done better with a one touch app.

    I have been a Linux user for many years (since about 1999) and have never had a memory issue in all that time (all nighters trying to get X working back in the day is another story).
  12.    #72  
    Quote Originally Posted by corbintechboy View Post
    What should be done is make a kill script that can kill all processes that are outside of the scope of a normal startup.

    Linux is very good at managing memory, better then Windows in every sense. If we are closing apps and they are still remaining in memory, this is an app issue.

    I could write a script, a better option would be some of our fine devs to make an app that can do this. A phone reset is nothing more then returning the device to the "start" state. This could be done with a script with no issue, done better with a one touch app.

    I have been a Linux user for many years (since about 1999) and have never had a memory issue in all that time (all nighters trying to get X working back in the day is another story).
    The apps all run in a single process, the same one that runs the UI.

    The Java services all run in a single process, the same one that runs the email, messaging, etc services.

    Apart from WebOS Internals C-based services (which are all written by people who know how to track down and fix memory leaks, and the ones I personally write are written to never call malloc at any time after startup), there's not much else to kill at a Linux process level.

    -- Rod
    WebOS Internals and Preware Founder and Developer
    You may wish to donate by Paypal to donations @ webos-internals.org if you find our work useful.
    All donations go back into development.
    www.webos-internals.org twitter.com/webosinternals facebook.com/webosinternals
  13. #73  
    Quote Originally Posted by rwhitby View Post
    The apps all run in a single process, the same one that runs the UI.

    The Java services all run in a single process, the same one that runs the email, messaging, etc services.

    Apart from WebOS Internals C-based services (which are all written by people who know how to track down and fix memory leaks, and the ones I personally write are written to never call malloc at any time after startup), there's not much else to kill at a Linux process level.

    -- Rod
    Then it must be a result of the way Linux uses memory (really not good on this type of device).

    Unused memory is wasted memory on a computer, but on a phone not so. If the phone uses the standard way of Linux thinking then closing apps leaves them cached in memory for later use. Linux has the power to free memory on demand but if the OS is seeing a high use then the OS simply will not allow a launch in order for Linux to make the memory available for said app.

    This is a flaw in using Linux in a device with such a restraint. Change the way Linux caches memory and problem solved. This is an area where I have no idea. Linux memory use is great, just not on our device for the above reason.

    EDIT: Just found this "sync; echo 3 > /proc/sys/vm/drop_caches"

    That may work to clear cache.
    Last edited by corbintechboy; 06/21/2010 at 10:37 PM.
  14. #74  
    Quote Originally Posted by cobrakon View Post
    +1000000

    We still stand behind our findings that poor coding is the reason for TMC errors. Unfortunately the constant new profiles and $$ for apps has become prohibitive for our cause....but our findings remain consistent that on a factory fresh Pre (sprint), you can run ANY number of "Stock" apps and be able to run the 3D game of your choice, close it, run any additional number of stock apps, and run the 3D game again without issue.

    Once you begin to introduce 3rd party apps, you will most certainly begin seeing TMC's. The main issue is not "outing" those devs who just may not know, but also having technical evidence to present. (Maybe we'll just upload all our data to WebOSInternals )
    While I'm inclined to agree with some of your post I'm not convinced that it is entirely 3rd party apps and poor coding. For example the browser while in itself may not have a memory leak it is one of the common factors leading to TMC. Just load up any one of many pages and not to long you will get TMC even if you freshly booted the device and the browser is the only app to have been used prior to TMC. Many users experience TMC, myself included (without a custom kernel), with a freshly booted device and during play of a 3D game. Some can't even open some 3D game after a fresh reboot because they would get the TMC with no other apps open.

    I personally have tested assassin's creed and with fair confidence can state it has a memory leak. How rapid the leak varied from as short as a minute or two to as long as 15 or 30 minutes. Changing to the current or just previous version of the uber kernel and due to the bug fix (details posted elsewhere in this thread I believe) they applied for TMC and compcache you can play assassin's creed for 2.5 hours before TMC occurs from the leak. Without compcache it was about 2 hours before TMC occurs from the leak. At the point of TMC performance degrades as the leak continues to slowly fill up memory. Bottom line is no matter when so far inevitably playing it will result in TMC. So far no other single app, combination of apps used, or extent of use of anything has resulted in TMC. Except the browser example I gave. The only other thing that cause TMC I confirmed to have been a result of the bug fix I mentioned was done to palm's kernel in the custom kernel. Considering the bug fix I have not under normal use of the browser been able to get TMC again. I'm pretty sure that I would have to use it a lot more to cram up more memory from leak(s).
    As requested: for my works on webOS patches and apps. Twitter: @larryboytw Patches: Small icons browser start page, 5x5 launcher. I have an AAS CIS Programming degree. I enjoy working on open source projects and alpha and beta testing.
    http://install.preware.org/ for easy to get up and running for patches and apps.
  15. #75  
    ROD and UNIXPSYCHO: Question:

    I sit at work most of the day and read PreCentral in awe of the skills and knowledge you and others have (NOTE: the beginning of this thread and all the code). I tinker around with Linux (Ubuntu), the innerworkings of the Pre and other stuff like HTML 5, CSS, etc. but my miniscule skills pale in comparison to what you all know.

    Could you please PM me (or list on here) where you got your training? What is it in? Where should I start? What order should I go in to be versatile in the "code" as you all seem to be? Is it Linux/Unix programming skills, C, C++, what? I really would like to know how you all got your start and are where you are today so I can make similar choices. I truly appreciate and admire the work you do and would love to walk a similar path.
  16. #76  
    Quote Originally Posted by StoneRyno View Post
    While I'm inclined to agree with some of your post I'm not convinced that it is entirely 3rd party apps and poor coding. For example the browser while in itself may not have a memory leak it is one of the common factors leading to TMC. Just load up any one of many pages and not to long you will get TMC even if you freshly booted the device and the browser is the only app to have been used prior to TMC. Many users experience TMC, myself included (without a custom kernel), with a freshly booted device and during play of a 3D game. Some can't even open some 3D game after a fresh reboot because they would get the TMC with no other apps open.

    ........
    Well my "dev team" (all 3 of us. lol) and I have been doing some fairly intensive testing on Sprint Pre's trying to determine the starter of the TMC phase that we are all in now. This has involved literally 100's of trips to the Doc (across all 4 phones) and countless Palm profiles created (thank goodness for 1200 email addys on our host!) in order to test. Our findings point that on all 4 stock Pre's (no 3rd party apps or 3PA's) we get the TMC at around 8-12 apps (avg) depending in the apps.

    The next test was to Doc the pre's and start buying *only* 3D games from the catalog and test repeatedly (thus buying the same games over and over) as this is the "control" for the testing to ensure starting from the same setup. We have been able to run ALL of the stock apps in any order and launch the 3D games (Glyder2, Asphalt, NFSU) repeatedly with nary a TMC in sight.

    The interesting part came about as we started performing the tests adding 3rd party apps into the mix (no homebrew on that phase) and the TMC started occuring in as few as 4 apps loaded. The difficult part is isolating which apps actually cause the problem by hitting the Doc and installing the 3PA's in different configurations to replicate the issue. We haven't "given up" per say but even with the "small" number of WebOS apps available, this has become a daunting task with 2544^2544 possible app combinations.

    We'd like more Pre's, but the 4 Pre's we are using have varied enough build dates that we feel confident in ruling out device anomalies.

    Rod's data is very interesting and it would be cool if we could somehow see those settings in play as we tested since that would likely truly reveal the "culprit" apps. Please understand we are not knocking any Devs but they could be inexperience issues cause their code to inefficiently use resources. (You see this in Sourceforge apps in Windows all the time)
  17. #77  
    cobrakon,

    Looks like you guys got a good method going on but I got a question. When you say TMC started occurring with as few as 4 apps, do you mean performing the identical test that consisted of 12 app no longer the same 12 apps can load without getting TMC (even after a reboot)? In other words it isn't random apps being picked making it seem like there is an issue? Some apps are memory hogs due to the nature of the app. So in that kind of app code efficiency is important to reduce both memory and processing usage. In fact I would even volunteer my services to free apps to clean up code some and hire out my services to paid app devs who need code cleanup. But in many of the simple apps it isn't as big a deal except for power users who are going to have 15+ app open at the same time. But what I was trying to drive at is if something is causing the same 12 app to not be able to be open at the same time that is something that should be a focus to find the cause.

    For example the most cards I have had open at the same time was maybe 8 or 10. I have never seen a need to do more than that even if I were to use multiple browser sessions I can't imagine having more than 10 cards open. But this isn't to say that one shouldn't be able to have 20 even 30 cards open if they want to. No matter how many new 3rd party apps I add or patches and modifications I make to my device I have consistency in which apps make up the most cards I ever have open at once. One thing that is clear to me is the bug fix previously discussed. With it in place I can now load up 6 or 8 heavy duty apps at the same time and the device run as smooth as with just a single app opened. Previously I would not be able to have other stuff open with a 3D game or would have to close everything but the browser to visit some web sites. Now I can have two browser cards pointed to sites that previously forced me to have to close cards and several other apps running at the same time, and experience no impact to performance.
    As requested: for my works on webOS patches and apps. Twitter: @larryboytw Patches: Small icons browser start page, 5x5 launcher. I have an AAS CIS Programming degree. I enjoy working on open source projects and alpha and beta testing.
    http://install.preware.org/ for easy to get up and running for patches and apps.
  18. #78  
    Please everyone, read what I posted.

    The issue is coming from the way Linux handles memory. I have been using it to many years to not understand. The issue is to many apps being cached in memory.

    Write a flush cache app using something like the example I posted above. This should in theory solve the issue.

    Pre looks to see how much memory is available, sees not much. Linux is amart enough to flush some cache to make memory available for app. Pre is not programmed in a way that will let Linux do this. Pre sees memory low and will not allow Linux to flush the cached memory.

    It really is simple logic. Linux is very good with memory, the Palm devs wrote something into the Pre that does not allow Linux to manage memory the way it was made to.

    Even if this was caused by a memory leak, still we would close the app (kill) and the only thing that could remain a killer is the amount of memory that app cached.

    I can run a top on my phone and see any app I launch. Can also close an app and see it gone. So everything runs in a process and therefor can be killed and does leave something in memory. TMC is a fault of the Pre OS for not letting Linux manage memory it was made to do.

    If someone wants to rewrite the way Linux handles memory, this could be fixed. Otherwise a simple cache cleaner app would solve the problem. One could even write something that wipes cache on any app close. This would make launching the same process again a little slower but the benefit being that there is more memory.

    Not the way Linux was made but a solution none the less.
  19.    #79  
    Quote Originally Posted by corbintechboy View Post
    I can run a top on my phone and see any app I launch. Can also close an app and see it gone. So everything runs in a process and therefore can be killed and does leave something in memory. TMC is a fault of the Pre OS for not letting Linux manage memory it was made to do.
    Have you actually run top on a webOS device? (The real Linux top, not JSTop).

    I suspect you have not, and are basing your assertions on usage of desktop Linux operating systems or perhaps other Linux-based mobile operating systems, but not webOS. If you are seeing new processes appear in top when you launch an application in webOS, then you must be using a different version of webOS from the 10 or 12 different versions that I have used over the last year.

    Please re-read post #73, which explains how apps (cards) operate in webOS. I will repeat again that they are not separate processes. The reason for this is explained in the "WebOS Internals" video from the recent Palm developer day. You may like to view that video for more information about exactly how memory is handled in webOS.

    -- Rod
    WebOS Internals and Preware Founder and Developer
    You may wish to donate by Paypal to donations @ webos-internals.org if you find our work useful.
    All donations go back into development.
    www.webos-internals.org twitter.com/webosinternals facebook.com/webosinternals
  20. TomD's Avatar
    Posts
    908 Posts
    Global Posts
    1,047 Global Posts
    #80  
    Quote Originally Posted by corbintechboy View Post

    EDIT: Just found this "sync; echo 3 > /proc/sys/vm/drop_caches"

    That may work to clear cache.
    Tried "drop_caches" in Terminus with Dev Mode and got"

    "Permission denied"

    Any ideas?
    Tom
    Pilot Prō --> M100 --> Trēō 600 --> Trēō 700p -- > Prē
Page 4 of 6 FirstFirst 123456 LastLast

Posting Permissions