Re: RS6000: Missing EABI symbols at link time

[Geoff Keating <geoffk at geoffk dot org>]

On 06/02/2006, at 6:30 AM, Daniel Jacobowitz wrote:

> On Mon, Feb 06, 2006 at 12:41:28PM +0000, Nick Clifton wrote:
>
> > Right - this is really the point.  The configure tests used by
> > libstdc++-v3 appear to include checking to make sure that the  
> > compiler
> > can compile and link an executable.  I do not know enough about
> > configure to know how to stop this link test and I was not sure what
> > gcc's policy actually was.  I guessed that the policy was that any  
> > gcc
> > compiler should be able to link together an executable without  
> > needing
> > any special command line options, or that if it couldn't do this,  
> > then
> > it would issue an informative error message telling the user what was
> > needed.  eg:
> >
> >   "gcc required one of -mfreebfd, -mads, ... to be specified on the
> > command line before it can create an executable."
>
> Or default to one of them (probably -msim)?

That's what the target powerpc-eabisim is for.  powerpc-eabi is  
specifically to prevent it defaulting to anything, and force the  
caller to make a choice, so that you can detect cases where, for  
instance, the libstdc++ configure process has broken.

It doesn't produce an 'informative error message' like that because  
the user is also allowed to supply their own crt* files, like

gcc mycrt0.o foo.c -o foo

in case one of the existing boards isn't what they want.

Attachment: smime.p7s
Description: S/MIME cryptographic signature