Re: fat binaries ?

Re: fat binaries ?

[H.J. Lu]

> 
> hallo,
> just a silly question out of curiosity:
> 
> are there plans to include the functionality to produce 'fat binaries' into
> the linux c compilers ?
> i mean the ability to automaticaly crosscompile for different linux platforms
> and bind everything into one binary? i think this would be very nice for
> commercial companies that will only provide binaries.
> 

I doubt if it will ever be supported directly. Maybe a front end?


H.J.

Re: fat binaries ?

[Mat Hostetter]

>>>>> "H" == H J Lu <hjl@innovix.com> writes:

    >>  hallo, just a silly question out of curiosity:
    >> 
    >> are there plans to include the functionality to produce 'fat
    >> binaries' into the linux c compilers ?  i mean the ability to
    >> automaticaly crosscompile for different linux platforms and
    >> bind everything into one binary? i think this would be very
    >> nice for commercial companies that will only provide binaries.
    >> 

    H> I doubt if it will ever be supported directly. Maybe a front
    H> end?

FWIW, I seem to remember one of NeXT's engineers saying at a NeXTWORLD
conference that NeXT had patented fat binaries "for defensive
reasons".  I haven't seen the patent, and my memory of that conference
is a bit fuzzy, so don't take my word for it.

-Mat

Re: fat binaries ?

[Samuel Figueroa]

If anyone is interested, you may want to look at what we did for our  
compilers for NEXTSTEP, which support the creation of 4-way fat binaries.  I  
doubt we'll have any time to send patches to FSF in a reasonable time frame,  
but we do make all our source code freely available, at least to our  
customers.

Re: fat binaries ?

[Richard Stallman]

I don't think it is essential to have fat binaries.
In fact I think it is more convenient to use other techniques
such as architecture-dependent symlinks together with separate
directories for the various architectures.

But if a convention is established for multi-architecture binaries, we
might as well support it.  All that would be needed would be a simple
program to extract and install subfiles in an archive of that kind.

This is a modular design which would not involve any changes in the C
compiler or GAS, and perhaps not even in the binutils.