Author Topic: Test build of dooble with qt5  (Read 139940 times)

mauro

  • Sr. Member
  • ****
  • Posts: 410
  • Karma: +3/-0
    • View Profile
Re: Test build of dooble with qt5
« Reply #255 on: December 27, 2023, 06:52:58 pm »
Thanks Dave for the new version - works great here too. And thanks Martin for requesting the password repository at www.github.com/textbrowser... looks like it may be in our future! Dooble just keeps getting better.

 @Mauro - I don't see your issue here either.

Hi David, test with www.repubblica.it and check

David McKenna

  • Hero Member
  • *****
  • Posts: 751
  • Karma: +24/-0
    • View Profile
Re: Test build of dooble with qt5
« Reply #256 on: December 29, 2023, 04:13:26 pm »
Hi Mauro,

  Yes, you are right, I see the same missing letters in the title on that website, both QT5 and QT6 Doobles - very strange. First I've seen that...

Regards,

TeLLie

  • Full Member
  • ***
  • Posts: 236
  • Karma: +12/-0
    • View Profile
Re: Test build of dooble with qt5
« Reply #257 on: December 31, 2023, 10:35:20 am »
Hello,
When I use dooble it takes minutes to scroll, I only have this with 2 laptops. I don't have this problem on my desktop comp.
Even if I have to fill something out, it takes minutes.
So I'm curious if others are experiencing this, I have had this since the first version of dooble or otter-browser. It doesn't matter whether I use version of Qt6 or Qt5.
Even when I create a new profile, scrolling remains very slow.
Strange thing is that I only have this with browsers, I don't see this with other gui programs.
That's why I still use FF as a browser.

Anyone have an idea?

mauro

  • Sr. Member
  • ****
  • Posts: 410
  • Karma: +3/-0
    • View Profile
Re: Test build of dooble with qt5
« Reply #258 on: December 31, 2023, 12:40:11 pm »
Hello,
When I use dooble it takes minutes to scroll, I only have this with 2 laptops. I don't have this problem on my desktop comp.
Even if I have to fill something out, it takes minutes.
........
Anyone have an idea?

Hi TeLLie, a very basic check : how is the cpu usage while a webpage is opened in Dooble ?
Look, if the answer is "99 to 100" that's the problem but I can't suggest a fix  ::)

TeLLie

  • Full Member
  • ***
  • Posts: 236
  • Karma: +12/-0
    • View Profile
Re: Test build of dooble with qt5
« Reply #259 on: December 31, 2023, 01:11:02 pm »
Hello,
When I use dooble it takes minutes to scroll, I only have this with 2 laptops. I don't have this problem on my de

Hi TeLLie, a very basic check : how is the cpu usage while a webpage is opened in Dooble ?
Look, if the answer is "99 to 100" that's the problem but I can't suggest a fix  ::)

Hi Mauro,

Cpu is up to 10%

David McKenna

  • Hero Member
  • *****
  • Posts: 751
  • Karma: +24/-0
    • View Profile
Re: Test build of dooble with qt5
« Reply #260 on: December 31, 2023, 01:41:58 pm »
 Hi Tellie,

 You say only on your laptops... what happens if you plug in a USB mouse and use it - is it still slow? Maybe there is a mouse driver issue with your mouse pad... I don't see it here with my laptops though. There is a test Amouse driver at ArcaNoae you could try - I needed it for one of my laptops for other reasons.

Regards,

TeLLie

  • Full Member
  • ***
  • Posts: 236
  • Karma: +12/-0
    • View Profile
Re: Test build of dooble with qt5
« Reply #261 on: December 31, 2023, 01:47:30 pm »
Hi Tellie,

 You say only on your laptops... what happens if you plug in a USB mouse and use it - is it still slow? Maybe there is a mouse driver issue with your mouse pad... I don't see it here with my laptops though. There is a test Amouse driver at ArcaNoae you could try - I needed it for one of my laptops for other reasons.

Regards,
Hi David
I use the latest test mouse from yesterday build.
I have a wireless mouse connected, but I could try with a wired mouse if that makes any diff.

Dave Yeo

  • Hero Member
  • *****
  • Posts: 4811
  • Karma: +101/-1
    • View Profile
Re: Test build of dooble with qt5
« Reply #262 on: December 31, 2023, 04:47:47 pm »
Hi Tellie, have you tried reducing the core count to see if any change?

TeLLie

  • Full Member
  • ***
  • Posts: 236
  • Karma: +12/-0
    • View Profile
Re: Test build of dooble with qt5
« Reply #263 on: December 31, 2023, 04:50:42 pm »
Hi Tellie, have you tried reducing the core count to see if any change?
Hi Dave
Yes I tried with 2 core same behavior
I will try with no cores and see..

Dave Yeo

  • Hero Member
  • *****
  • Posts: 4811
  • Karma: +101/-1
    • View Profile
Re: Test build of dooble with qt5
« Reply #264 on: December 31, 2023, 05:09:17 pm »
Hi Tellie, have you tried reducing the core count to see if any change?
Hi Dave
Yes I tried with 2 core same behavior
I will try with no cores and see..

MTRR problem? Look in a testlog for "shadow buffer". Here I get,
...
Code: [Select]
Video Aperture: Base=0xE0000000 Size=0x900000
Attempt to allocate shadow buffer...Success
Shadow Buffer locked at 0x5E116000, Length 0x900000 bytes
Trying E0000000 size 1000000.
MTRR Write Combine caching enabled Base=E0000000 Size=1000000.
...
Code: [Select]
Profile setting for Shadow...Enabled
...

TeLLie

  • Full Member
  • ***
  • Posts: 236
  • Karma: +12/-0
    • View Profile
Re: Test build of dooble with qt5
« Reply #265 on: December 31, 2023, 05:26:45 pm »
Hi Dave,
Video Aperture: Base=0x90000000 Size=0x800000
Attempt to allocate shadow buffer...Success
Shadow Buffer locked at 0xBE5F7000, Length 0x800000 bytes
Trying 90000000 size 800000.
MTRR Write Combine caching enabled Base=90000000 Size=800000.

and
Profile setting for Shadow...Enabled

Also with 0 cores i get same behaivor

David McKenna

  • Hero Member
  • *****
  • Posts: 751
  • Karma: +24/-0
    • View Profile
Re: Test build of dooble with qt5
« Reply #266 on: December 31, 2023, 06:27:58 pm »
Tellie,

  How about your settings - do you have --single-process set? How about 'Animated scrolling'? Might be worth comparing the settings on the desktop with the settings on your laptops...

Regards,
« Last Edit: December 31, 2023, 06:30:36 pm by David McKenna »

Rich Walsh

  • Sr. Member
  • ****
  • Posts: 339
  • Karma: +23/-0
  • ONU! (OS/2 is NOT Unix!)
    • View Profile
Re: Test build of dooble with qt5
« Reply #267 on: December 31, 2023, 07:55:11 pm »
When I use dooble it takes minutes to scroll, I only have this with 2 laptops. I don't have this problem on my desktop comp. Even if I have to fill something out, it takes minutes.

I have (or had) this problem. Answer these questions to see if we have the same issue:

1) Are these machines able to emit a beep? Not a sound through the soundcard but a tone through the speaker or transducer on the mobo. If yes, it's a different problem.

2) After scrolling and finding that nothing happens, does moving the mouse around _a lot_ over the window cause the accumulated scroll actions to happen all at once?

3) Assuming you're running AOS, run 'ACPISTAT | MORE'. At the top is a list or IRQs and the number of times they've been triggered. Is the count for IRQ 0 very low (in hundreds or less) or is IRQ 0 simply missing (either of these may be true)?

If the answer to (1) is no but (2) and (3) are yes, then we have the same problem: the clock which feeds the 8254 timer circuit has been gated off.

The 8254 has 3 timers: one is used to make sound while another is used by many apps such as Dooble as a "high-resolution timer". To use it, the app sets a counter to the number of milliseconds required. Each clock pulse then causes it to decrement that counter. When that counter reaches 0, it issues an IRQ 0.

If the clock signal has been turned off, the counter never decrements, the IRQ is never triggered, and Dooble fails to scroll. There is one oddity here: for unknown reasons, significant USB activity (or at least, USB mouse activity) will trigger an IRQ 0. That's why moving the mouse may cause scrolling - it mimics the timer going off.

I spent a huge amount of time identifying the nature of the problem but never came up with a solution, only a work-around. The high-res timer can only be used by one app at a time. If it's busy, the app has to fall back to one of the system's conventional timers. So, I wrote a little util that claims the HRT for as long as possible to force Dooble and other stuff to use an alternate timer. This works fine for scrolling but multimedia performance is terrible - VLC is unusable.

AFAICT, only a firmware update would fix the problem - and that's unlikely because MS stopped using the 8254 in Win 7 or 8. Consequently there's no way to demonstrate to some Tech Support droid running Win 11 that there even is a problem.

TeLLie

  • Full Member
  • ***
  • Posts: 236
  • Karma: +12/-0
    • View Profile
Re: Test build of dooble with qt5
« Reply #268 on: January 01, 2024, 12:48:56 pm »
When I use dooble it takes minutes to scroll, I only have this with 2 laptops. I don't have this problem on my desktop comp. Even if I have to fill something out, it takes minutes.

Hi Rich,

First all the best for 2024 ....

I have (or had) this problem. Answer these questions to see if we have the same issue:

1) Are these machines able to emit a beep? Not a sound through the soundcard but a tone through the speaker or transducer on the mobo. If yes, it's a different problem.
    Yes i hear a beep sometimes. Alto is a sound that i get when f.e running a app that i have compiled but it's gives a trap
2) After scrolling and finding that nothing happens, does moving the mouse around _a lot_ over the window cause the accumulated scroll actions to happen all at once?
   Yes
3) Assuming you're running AOS, run 'ACPISTAT | MORE'. At the top is a list or IRQs and the number of times they've been triggered. Is the count for IRQ 0 very low (in hundreds or less) or is IRQ 0 simply missing (either of these may be true)?

The version of ACPI.PSD that is installed is 3.23.17
The system is operating in Symmetric mode (Mode 2)
The kernel is 14.204_SMP
The retail PSD is installed

Number of interrupts available: 60
IRQ 00 count 14
IRQ 01 count 1049
IRQ 08 count 1137885
IRQ 09 count 11
IRQ 16 count 513
IRQ 17 count 27
MSI 55 count 8201
MSI 56 count 77012
MSI 57 count 18757

If the answer to (1) is no but (2) and (3) are yes, then we have the same problem: the clock which feeds the 8254 timer circuit has been gated off.

The 8254 has 3 timers: one is used to make sound while another is used by many apps such as Dooble as a "high-resolution timer". To use it, the app sets a counter to the number of milliseconds required. Each clock pulse then causes it to decrement that counter. When that counter reaches 0, it issues an IRQ 0.

If the clock signal has been turned off, the counter never decrements, the IRQ is never triggered, and Dooble fails to scroll. There is one oddity here: for unknown reasons, significant USB activity (or at least, USB mouse activity) will trigger an IRQ 0. That's why moving the mouse may cause scrolling - it mimics the timer going off.

I spent a huge amount of time identifying the nature of the problem but never came up with a solution, only a work-around. The high-res timer can only be used by one app at a time. If it's busy, the app has to fall back to one of the system's conventional timers. So, I wrote a little util that claims the HRT for as long as possible to force Dooble and other stuff to use an alternate timer. This works fine for scrolling but multimedia performance is terrible - VLC is unusable.

AFAICT, only a firmware update would fix the problem - and that's unlikely because MS stopped using the 8254 in Win 7 or 8. Consequently there's no way to demonstrate to some Tech Support droid running Win 11 that there even is a problem.

Is it from harm to try you're prog or?

Rich Walsh

  • Sr. Member
  • ****
  • Posts: 339
  • Karma: +23/-0
  • ONU! (OS/2 is NOT Unix!)
    • View Profile
Re: Test build of dooble with qt5
« Reply #269 on: January 01, 2024, 06:34:00 pm »
Is it from harm to try you're prog or?

I sent it to your GMail address in 'WPZ'  format which should get it past Google's malware filters. Let me know the result.