RE: [ros-svn] [ion] 19196: - Add new configuration option "SARCH" todefine the sub-architecture. Examples include i386-xbox, ppc-be/le (little/bigendian), or any other sub-types of the specified ARCH.

[Ge van Geldorp]

>   From: ion(a)svn.reactos.com
>  
>  - Add new configuration option "SARCH" to define the 
>  sub-architecture. Examples include i386-xbox, ppc-be/le 
>  (little/bigendian), or any other sub-types of the specified ARCH. 
This means we will have to build, test and distribute different .isos for
different sub-architectures, only to save a few bytes in the
freeldr/setupldr binaries. I implemented calling of sub-architecture
specific code via a table of function pointers precisely to be able to keep
one binary (per architecture).
Please revert.

Gé van Geldorp.

[Alex Ionescu]

Ge van Geldorp wrote:

> >  From:
> > ion(a)svn.reactos.com
> >
> > - Add new configuration option "SARCH" to define the 
> > sub-architecture. Examples include i386-xbox, ppc-be/le 
> > (little/bigendian), or any other sub-types of the specified ARCH.
> >     
> >  
> This means we will have to build, test and distribute different .isos for
> different sub-architectures, 
>  
That day will come one day or another.

>  only to save a few bytes in the
> freeldr/setupldr binaries.
>  
More then "few", and it wasn't only for Freeloader. I am a man of 
vision; I thought of this and its long-term effects. I'm sure they'll 
become apparent in a year or two when we start supporting other 
architectures and sub-architectures.

>   I implemented calling of sub-architecture
> specific code via a table of function pointers precisely to be able to keep
> one binary (per architecture).
>   
>  
Not that I mind the design (which is very well made) but there as many 
"anti-"FAT"-binaries" arguments as there are 
"anti-segregated-binaries"... and I don't want to start a war on which 
design ReactOS is going to support. But since you did most of the recent 
work on Freeloader and the XEN/XBOX ports, I consider you Freeloader's 
current maintainer and your choice is final to me.

>  Please revert.
>   
>  
Done, thanks for the clear communication on your disagreement.

>  Gé van Geldorp.
>
>   
>  
Best regards,
Alex Ionescu

>  _______________________________________________
> Ros-dev mailing list
> Ros-dev(a)reactos.org
> http://www.reactos.org/mailman/listinfo/ros-dev
>
>   
>

[Thomas Weidenmueller]

Alex Ionescu wrote:
>   That day will come one day or another. 
How many so totally incompatible subarchitectures are there? I liked the
way we did it before, not having to have to build multiple versions for
similar architectures.

>   More then "few", and it wasn't only for
> Freeloader. I am a man of
>  vision; I thought of this and its long-term effects. I'm sure they'll
>  become apparent in a year or two when we start supporting other
>  architectures and sub-architectures. 
I wouldn't consider the xbox as a subarchitecture, detecting it on the
fly imo is much better than having to build separate binaries. They're
not totally incompatible. Although the idea of subarchitectures is a
good one, i'd really just use it for "real" subarchitectures.

Just my two cents...

- Thomas

[Thomas Weidenmueller]

Ge van Geldorp wrote:
>   This means we will have to build, test and distribute
> different .isos for
>  different sub-architectures, only to save a few bytes in the
>  freeldr/setupldr binaries. 
I don't think it was about saving a few bytes because actually the size
of the binary increased by 4 kb (at r19202).

- Thomas