[Openswan dev] Re: [uml-devel] Re: [Openswan CVS] openswan-2/linux/net/ipsec addrtot.c,1.16,1.17

[BlaisorBlade]

On Monday 20 September 2004 06:45, CaT wrote:
> On Sun, Sep 19, 2004 at 11:53:28PM -0400, Michael Richardson wrote:
> >     CaT> In include/asm-i386/string.h for example. Not all arches have
> >     CaT> it defined but a few do.
> >
> >     CaT> With the change btw, the kernel compiled.
> >
> >   Using __HAVE_ARCH_STRSTR doesn't work on UML.
> >   It appears that the strstr.c from arch/i386/lib/strstr.c doesn't get
> > compiled. I tried to get it compiled, but I seem to be missing some
> > symbol table magic. (Adding EXPORT_SYMTAB(strstr) to the file didn't
> > help).
> >
> >   Any ideas on this?

> I'm not expert (seriously - this isn't just modesty :) but it looks
> like that uml is undefining strstr but not __HAVE_ARCH_STRSTR so that
> when it's compiled on an arch that defines __HAVE_ARCH_STRSTR because
> it has strstr it's only half undefined.

> If I'm right it appears to 
> be a bug in the part of uml.
Ok, that's not a bug, but one UML choice: it will use the host libc version of 
strstr. 

> Check linux/arch/um/os-Linux/user_syms.c.
The problem is just the fact that "EXPORT_SYMBOL(strstr)" is commented out in 
THAT file. If you compile the code which uses that statically, it will work, 
so you're not on the right road.

> It includes string.h 
> which includes asm/string.h which includes asm/arch/string.h which
> points to asm-i386/string.h which would define the above.

> I've CCed the happy-fun folks at UML to be sure because my feeling that
> I'm talking sheer bollocks is about as strong as my feeling that I am
> not. :)
I don't know what "sheer bollocks" means, but I guess this is the case...

Bye
-- 
Paolo Giarrusso, aka Blaisorblade
Linux registered user n. 292729