[ros-dev] CRT source code issue

Aleksey Bragin aleksey at reactos.org
Mon Aug 6 14:15:52 CEST 2007


Yes, the whole issue is at pushing operands to stack, performing well- 
documented assembly instructions to perform the asked action (div,  
mul, shl, shr, etc), and handling some special, though obvious, cases  
(like if you do a shl of some longlong number by more than 64 bits,  
you always end up with 0, there is no need to call math instructions  
in order to compute this).

I will let you know results when I speak with SDL.


WBR,
Aleksey Bragin.

On Aug 4, 2007, at 8:52 PM, Daniel Hazelton wrote:

> On Saturday 04 August 2007 12:28:09 Aleksey Bragin wrote:
>
> I can't see where that file has anything to do with the functions  
> (e.g:
> sqrt()/alldiv()) that Betov claims are duplicated. (I've gone  
> through that
> file and can't find any single place where it does any more math than
> controlled bit-shifts. No advanced math or anything.)
>
> While there might be similar code, for, say, memory barriers, that  
> codes
> construction would be dictated by the specification of the target  
> machine
> (and, to some extent, the OS). If the code for sqrt/alldiv is the same
> between SDL/ROS I don't see a problem - they are both GNU projects  
> and "cross
> pollination" like that is legal.
>
> Whats more is that, while assembly is abstract enough to let some  
> things be
> done in more than one way, most of the more complex functionality -  
> like math
> algorithms - are so limited in the possibilities of variation that  
> I can
> easily see two unconnected people arriving at the exact same  
> solution. QED:
> Even if the code appears identical (except for things like variable  
> names,
> et. al) there is a compelling reason for it and it does not mean  
> the code has
> been copied in any way, shape or form. (And if you look at the ROS  
> commit
> history I'm sure the code in question existed before MS released  
> the SSCLI2.0
> source. Which means that the code can't be an "illegal use of MS  
> property" -
> it was around before said property was available).
>
> ...Okay, enough of this outsider sticking his neck out. I've said  
> enough and
> it's time to disappear and wait for a really usable version of ROS  
> to hit.
> (though I have compared it to XP under Qemu and ROS rocks!)
>
> DRH


More information about the Ros-dev mailing list