gmp 4.2 fails tests on G5

[Daniel Lord]

Jack ,

So we now have valid builds on OS X PPC and i386 Now you know why I   
am asking that the GMP site be updated.
The GMP site is stating outright falsehoods  blaming Apple's  
assembler and compiler when it is nothing of the sort.
The site and the conclusions it states are blatantly and patently  
false and need to be fixed.


Daniel

On Apr 26, 2006, at 17:57, Jack Howarth wrote:

> Daniel,
>      I got a clarification of this issue on the darwin-dev mailing  
> list...
>
> ---------------------------------------------------------------------- 
> -----
> -fno-common is not the issue.
>
> The issue is that GMP 4.2 compiles with:
> using ABI="mode64"
>        CC="gcc"
>        CFLAGS="-m64 -mcpu=970 -fast"
>        CPPFLAGS=""
>        CXX="g++"
>        CXXFLAGS="-m64 -mcpu=970 -fast"
>        MPN_PATH=" powerpc64/mode64 powerpc64/vmx powerpc64 generic"
>
> If one reads the documentation, one finds that -fast implies - 
> mdynamic-
> no-pic, which generates object files that are not suitable for use in
> a dylib. The gcc man page has:
>
>             To build shared libraries with -fast, specify -fPIC on
> command
>             line.
>
>
> Shantonu
> ---------------------------------------------------------------------- 
> ----
> So Darwin builds of gmp should be appending -fPIC after the -fast on
> the compile line.
>                   Jack

Daniel Lord
daniellord at mac.com
---
"You will never regret getting up too early,
and you'll always regret getting up too late,
but sometimes you may regret giving up too late."
                           -- Mountaineer's Adage