Re: fat binaries

[christos%astron.com@localhost (Christos Zoulas)]

In article <B090D4F3-D964-49F8-8F61-0BC3C6E61D49%ieee.org@localhost>,
Gregory McGarry  <g.mcgarry%ieee.org@localhost> wrote:
>I really like the concept of the mach fat binaries and have always  
>thought that NetBSD could benefit from their use more than any other  
>operating system.
>
>While NetBSD supports fat binaries containing macho executables, I  
>did a proof-of-concept to put native ELF executables inside the fat  
>container.  I wrote a lipo(1) tool based on the same tool on OS X.
>
>You can find patches at http://ftp.netbsd.org/pub/NetBSD/misc/ 
>gmcgarry/fat.  It's a grotty hack, but it proves that the concept is  
>not difficult.  The diff is against an 8-month-old version of - 
>current.  The cat.fat binary supports amd64, i386 and arm.
>
>What's the general feeling about such a feature?  I think it would be  
>great for universal boot cds.  It would be nice to boot a "netbsd  
>desktop" between an amd64 and i386 kernel using the same userland.
>
>Further changes would be required to the toolchain before widespread  
>adoption.  gdb works, but will only recognise the ELF binaries  
>extracted from the fat binary.  ld would need to grok shared libs  
>inside the fat container.
>
>If there is enthusiasm for the feature, I'll try to factor common  
>code between ELF and macho for a clean patch.  I'll also look into  
>modifying pcc to build fat binaries.

Perhaps it is better to do this using something like HP/UX's CDF's?

http://www.informatik.uni-frankfurt.de/doc/man/hpux/cdf.4.html

christos