Hi Ros Arm,
you think you got a good case. No you don't.
Yes, I decided to implement an easy function. And I decided to implement HalHandleNMI because I had a Computer that had a broken memory module. I could run DOS for hours but trying to install Windows NT4 always caused an NMI bluescreen.
Your magic constants argument doesn't work as well as my source of information is a number of articles in the german computer magazine C'T from 1988/1989 about the hardware component of an IBM-compatible Personal Computer (PC/XT/AT and IBM PS/2). C'T issue 7/1988 includes an article 'Die PC-Bausteine'(The PC-Components) on the pages 164 to 171 about the PIOs 8255 in the PC/XT and its successors in the AT and PS/2 which implement the ports 0x60 to 0x64. In this article you can find a detailed description about masking the NMI and detecting the source of an NMI. This article also includes tables that describe the function of each bit of the PIO port. The author doesn't provide any symbolic constants for the bit masks so I decided to use the hex values instead.
The call to the function KeEnterKernelDebugger was added intentionally because if neither bit 7 nor bit 6 is set upon an NMI the 80x87 coprocessor is the source of the NMI. In this case you might want to debug your application.
If you like, I can send you picture of page 166 where you will find the tables I mentioned above.
Now about the order of the two if-statements in HalHandleNMI. Well with only two of them you have a 50 percent chance choose the 'right' sequence. OTOH, I knew about the memory parity error first as I had seen it on my own machine. And I guess a friend told me about the second message. Perhaps he got used a text editor or the strings.exe utility from sysinternals.com. I don't remember that any more as this happened almost 10 years ago.
According to your train of thougt a 600 line file is not protected by copyright law if it was written by 100 programmers where each of them contributed exactly 6 lines of code. That might be true for your country but in Germany we dont have minimum number of lines. It depends on the context. A slogan like President Obamas "Change." can be protected by copyright.
To make a long story short, you can change the license of the file if you get permission by all authors and replace my implementation of HalHandleNMI by your own one.
Regards Eric
Ros Arm wrote:
Hi Eric,
We apologize for the misunderstanding. The relationship of that field was confusing, since some files have authors that don't correlate to SVN history at all.
In relation to what eVb was saying, I have looked up the entire SVN history of the misc.c file, which originally started out in NTOS.
Your one and only actual contribution is here:
http://svn.reactos.org/svn/reactos/trunk/reactos/ntoskrnl/hal/x86/misc.c?rev...
Revision 1026, almost 10 years ago.
In that revision, you created the file and "implemented" HalHandleNMI.
It is interesting that your implementation of HalHandleNMI happens to print out the exact same messages that the NT4 HAL dumps on the screen.
To obtain the same messages "clean-room", you must've been able to generate all the possible NMI traps on your machine, which is unlikely.
Also, the choices of "0x40", "0x20" and "0x61" appear to be magic constants, instead of properly documented constants referring to the NMI status port and the NMI status bits.
Additionally, the same "KeEnterKernelDebugger" call is present both in the NT4 HAL and your implementation of this function, even though, as of revision 1993:
http://svn.reactos.org/svn/reactos/trunk/reactos/ntoskrnl/kd/kdebug.c?revisi...
That function only printed a string -- so there appears to have been little reason to make this API call other than to duplicate the original NT HAL's behavior.
Furthermore, why even implement this function at all back in 2000? There was no NMI task gate back then, no NMI handler, and no caller of HalHandleNMI -- therefore, this function was impossible to test in ReactOS. It seems to have been implemented just for the sake of implementing "an easy function", without any actual real use.
Whether this function, in 2000, was by coincidence an exact replica of the NT4 HAL HalHandleNMI, and you happened to have experienced all possible NMI traps and their messages on the screen, and you decided not to document the port and status bit values, or whether it was reverse engineered does not change the fact that the code was 100% identical. Therefore, I believe this was eVB's point that the true "copyright" of this function belonged to nobody else but Microsoft, and so the PROGRAMMERS field seemed logical to include the maintainer, not the "copyright" owner.
So before lambasting about copyright ownership and licenses, perhaps you should be more careful.
It is an entirely different argument whether or not there is any other way to do this -- clearly, by now, it is known what NMI messages the Windows STOP screen will display, and ReactOS must display the same NMI messages. Since both Windows NT and ReactOS use the same routine for printing to the screen, it also makes sense that they'd both be using the same HalDisplayString API. One could argue that since you'd probably want to debug the NMI, it might make sense to enter the kernel debugger, which, since it is an exported kernel function that ReactOS must implement, ends up being the same call as Windows. So whether it was clean roomed, copied, reverse engineered, or fabricated out of thin air to match this specification, the resulting code would look the same (however, it should be better documented so the values don't look like magic). Of course these are all rational arguments that can be made *after* the function has been written, the burden of proof as to IF this is
how the function was already written is what matters.
There are many questions as to whether or not those 6 lines of code are actually copyrighted at all, whether you or anyone else has any rights to those 6 lines of code since they are identical to Windows (by necessity), and whether or not the addition of ~460 additional lines to misc.c and improvement of the original function still enable you to refuse a GPL->BSD license change solely based on 6, 10-year-old, lines of code that are identical to an even older copyrighted source base.
To push this analogy further, I could decide that the following piece of code
"printf("%s\n","
is GPL, and block anyone from ever using it in a module that is BSD.
I think even GPL requires at least 10 lines of code before being able to claim the whole module is GPLed (and it's probably more complicated than that).
So to close the topic: it is unclear if you have any ownership of those original 6 lines of code, and it is even further unclear if 6 lines of interface code in a 466 line module allow you the right to dictate the license of the module.
-r
Ros Arm wrote:
The "ownership" and "copyright" of this code is not belonging to either you or anyone else but the original source it came from.
The PROGRAMMERS field indicates who is responsible for a piece of code in this project, is it not? In this case, since you have quitted many years ago, it makes sense this code is now belong to us.
It is Linux concept of "maintainer".
Thanks you, [eVb]
Hi Ros Arm,
you are totally wrong here. The PROGRAMMERS field could be called AUTHORS or CONTRIBUTORS and lists those who claim copyright or authorship to a file. So removing someone from the PROGRAMMERS field is a very disrespectul act, is a copyright violation and must be reverted as soon as possible.
You can claim sole authorship to a file if you created the initial version of a file.
You can also claim authorship to a file if you modified it in a significant way. In this case you can ADD your name to the END of the list of programmers. So the author who wrote the first version should always stay on top of the list. Changing a "+" to "-" or adding "+ 1" is not enough to claim authorship.
And in no case you have the right to change the license of an existing file without having prior permission to do so by ALL authors of that file. As far as I can remember, I did NOT grant permission to relicense my code under the BSD license. My contributions to ReactOS were, are and will be published under GPL or LGPL only.
Therefore I ask you to revert the following patch: [ros-diffs] [ros-arm-bringup] 44860: NMI Support Patch 6: [HAL]: Fix NMI recursion issues. [HAL]: Reset the display during NMI and paint the NMI Screen of Death.
In case you fail to revert the patch within a week (aka 7 days) I will revert this patch myself and restore the previous version.
And by the way, DON'T INTRODUCE YET ANOTHER FUCKIING CODING STYLE!!!! Use the coding styles that are already being used!!!!
Oh, and before I'm done: I don't like to talk to meaningless nicks like Ros Arm, RosLeg or ROSIDONTKNOWHOWTOFUKINGCALLMYSELF. Reveal your identity! I don't see a reason why you have to hide behind you nick. If you need to hide your identity you should better leave ReactOS because these reasons can only endanger the project. And I will always prefer to throw a single contributor out of the team rather than endangering the whole project.
And finally, my name is Eric Kohl and I joined the ReactOS project in December 1998. So I guess I have the right to voice my opinion here.
Regards Eric
Ros-dev mailing list Ros-dev@reactos.org http://www.reactos.org/mailman/listinfo/ros-dev
Ros-dev mailing list Ros-dev@reactos.org http://www.reactos.org/mailman/listinfo/ros-dev