Page 1 of 2

r55833

Posted: Tue Feb 28, 2012 11:34 pm
by DOSGuy
Whatever changes were made to win32k.sys in r55833, ReactOS will no longer install in Bochs 2.5.1 as of that version. Immediately after the ROS startup screen following stage 1 installation, I BSOD with the following message:
win32k.sys

PAGE_FAULT_IN_NONPAGED_AREA

win32k.sys - Address FCACF774 base at FCA11000, DateStamp 4f47b169
Subsequent boots just go to a black screen after the startup screen.

Apparently some progress is being made on fixing this, as things are getting a little better by r55908. I still BSOD after the startup screen following stage 1 installation, but when I reboot, ROS moves on to second stage installation, which completes successfully.

I bring this up because Bochs usually performs similarly to real hardware, but I haven't seen anyone mention getting a BSOD during installation, and this has been going on for 75 commits. Is no one else having this problem?

Re: r55833

Posted: Wed Feb 29, 2012 12:01 am
by Bblaauw
Not tried this, sorry. I assume you got "log: -" set in your bochsrc to get visible output? Or specify a textfile there to log to and file a bug.
Most people are likely using VMware, VirtualPC and QEMU as they're all a bit faster than Bochs.
For the record, I've not been able to boot ReactOS in real hardware either for a while. Initially LiveCD got to desktop, now hangs on USB stuff. Not sure if it's related in some way to the issues you mention here.

Re: r55833

Posted: Wed Feb 29, 2012 12:05 am
by Z98
LiveCD is nowhere as religiously tested as the BootCD. 90% of the time devs don't bother checking it. You'll need to hound them fairly heavily if you want them to deal with livecd issues quickly.

Re: r55833

Posted: Wed Feb 29, 2012 12:40 pm
by gabrielilardi
Z98 wrote:LiveCD is nowhere as religiously tested as the BootCD. 90% of the time devs don't bother checking it. You'll need to hound them fairly heavily if you want them to deal with livecd issues quickly.
Then clearly 90% of the devs ignore masses taste:

Image

Re: r55833

Posted: Wed Feb 29, 2012 3:08 pm
by vicmarcal
WB gabriel ;)

Re: r55833

Posted: Wed Feb 29, 2012 8:21 pm
by FlyingIsFun1217
What about nightly builds?

FlyingIsFun1217

Re: r55833

Posted: Thu Mar 01, 2012 5:54 pm
by DOSGuy
And now things are getting worse. As of r55943, I get a BSOD after Stage 1 installation, and then on every subsequent boot, whereas r55908 would continue to Stage 2 after the first BSOD. I get the same thing in VirtualBox, so I'm running out of platforms to test ReactOS with.

Re: r55833

Posted: Thu Mar 01, 2012 9:14 pm
by vicmarcal
DOSGuy wrote:And now things are getting worse. As of r55943, I get a BSOD after Stage 1 installation, and then on every subsequent boot, whereas r55908 would continue to Stage 2 after the first BSOD. I get the same thing in VirtualBox, so I'm running out of platforms to test ReactOS with.
Well..this is called "Development". Big changes are been introduced to ReactOS (newCC, Memory manager rewrites,etc...)...Keep sending debuglogs and bugreports to Bugzilla ;)
Are they constant bugs?or happening from time to time?
You have to post the VM settings plus the Specific revision...as far as I know ReactOS Testbots are working, which means we can run their ISOs in Virtualbox and Qemu without problems.

Re: r55833

Posted: Thu Mar 01, 2012 9:48 pm
by DOSGuy
The great thing about Bochs is that it isn't a VM. Until r55833, it just worked. I'll screw around with the VM settings in VirtualBox to see if I can get it working, but it won't install with the default options for Windows 2003.

Edit: As of r55952, I can get to Stage 2 installation after the first BSOD in Bochs again.

Re: r55833

Posted: Fri Mar 02, 2012 1:24 am
by BrentNewland
DOSGuy wrote:The great thing about Bochs is that it isn't a VM.
Bochs is indeed a Virtual Machine; it just emulates more hardware than other Virual Machines.

Re: r55833

Posted: Fri Mar 02, 2012 10:09 am
by wildschwein
Look at these big kernel changes with 55938
http://svn.reactos.org/svn/reactos?revi ... w=revision

I am no dev, i have no plan about ros internals :-) but i assume, that this "big" chances can make trouble (but if they work, they are very informative about the place of bugs)

Re: r55833

Posted: Fri Mar 02, 2012 5:16 pm
by DOSGuy
BrentNewland wrote:Bochs is indeed a Virtual Machine; it just emulates more hardware than other Virual Machines.
While it's true that every emulator is a "virtual machine", when people talk about VMs, they're almost always talking about virtualization. Bochs is just an emulator, it does not use virtualization. Virtualization is a way to allow a guest operating system to use your computer's actual hardware, whereas an emulator must emulate all of a system's hardware, including hardware that your system actually has. On an x86 computer, Bochs must still emulate an x86 CPU (even though you have one), whereas VirtualBox doesn't have to emulate an x86 CPU because it can use the one that you have in your computer. This makes VirtualBox much faster than Bochs, but VirtualBox was hit hard by the "mshtml bug" and Bochs wasn't. I like using Bochs because it almost always works, which I think has something to do with the fact that it uses emulation instead of virtualization.

Re: r55833

Posted: Mon Mar 05, 2012 3:28 am
by milon
I just want to add that it's not to say that VBox is inaccurate - real hardware was also affected by the infamous "MSHTML bug". Bochs & VBox both have their pros & cons.

Re: r55833

Posted: Mon Mar 05, 2012 6:42 pm
by DOSGuy
milon wrote:I just want to add that it's not to say that VBox is inaccurate - real hardware was also affected by the infamous "MSHTML bug".
Indeed, and VBox suffered from it for probably the same reasons as real hardware did. Bochs didn't, which I appreciated.

It looks like the recent NewCC commits have solved the problem. As of r56034 I'm able to install ROS without any BSODs in Bochs 2.5.1!

Re: r55833

Posted: Mon Mar 05, 2012 8:28 pm
by EmuandCo
The fix was committed before the newcc fixes and additions, check the commit messages from about two weeks ago.