Re: Cross solib support; multi-arch problem?

[Kevin Buettner <kevinb at redhat dot com>]

On Nov 7,  1:37pm, Andrew Cagney wrote:

> >> So, I added a tm-cris.h file containing:
> >> 
> >> #define GDB_MULTI_ARCH 1
> >> #include "solib.h"
> > 
> > 
> > Unfortunately, I think this is going to be needed for the medium term. 
> > In the fullness of time, we'd like to get all of the shared library
> > support consolidated to using only solib.c for the interface to the
> > rest of GDB.  Once that is done, ADD_SOLIB and company won't be need
> > to be macros any more.  To be more precise, the shared library
> > framework present in solib.c will be present for all targets whether
> > it's needed or not.  When SOLIB_ADD() (or probably just solib_add())
> > is called, it'll simply not do anything when the target in question
> > doesn't have/require shared library support.
> 
> Kevin, what exactly does it take to do this?  Can the GDB group grit 
> their teath and throw a switch?

Maybe.  As noted in another thread, I think the following files will
need to be converted:

    coff-solib.c
    irix5-nat.c
    pa64solib.c
    somsolib.c
    xcoffsolib.c

Hmm...  it looks like win32-nat.c needs to be added to this list too.

Anyway, to get an idea about the sorts of things that need to be done,
take a look at osfsolib.c and then look at how Nick converted it into
solib-osf.c.  I think converting the code will be the easy part.  The
hard parts will be to 1) find machines on which the above are needed
and 2) put the ports into good enough condition so that the converted
code can actually be tested and compared against the original.  

Kevin