Hi Pedro,

> Thanks!  This is definitely the way to go.

My pleasure :).

> > +   If CLOSE_LDINFO_FD is nonzero, then this routine also closes
> > +   the ldinfo_fd file descriptor.  This is useful when the ldinfo
> > +   data is obtained via ptrace, as ptrace opens a file descriptor
> > +   for each and every entry; but we cannot use this descriptor
> > +   as the consumer of the XML library list might live in a different
> > +   process.  */
> 
> How interesting.  So AIX's list is better than glibc in that we
> could use that descriptor to avoid the "library file gdb opens is
> no longer the same as the inferior opened" issue.

True. Now, if only we could send the "fd" across processes :-).

> > +struct ld_info
> > +{
> > +  int next;
> 
> I'm surprised this is signed.
> It was unsigned on Ulrich's version: http://sourceware.org/ml/gdb-patches/2007-05/msg00321.html

Right you are. I read sys/ldr.h pretty carefully, and was convinced
that they had defined it as an int, but not so. For the record:

 ==============================================================================
 | Field Name      | __ld_info32 size;type(s)| __ld_info64 size; type(s)      |
 |============================================================================|
 | ldinfo_next     | 4; uint                 | 4; uint                        |
 | ldinfo_flags    | n/a                     | 4; uint                        |
 | ldinfo_fd       | 4; int                  | 4; int                         |
 | ldinfo_fp       | 4; struct file * / uint | 4; struct file *      / uint   |
 | ldinfo_core     | 4; int           / int  | 8; long long          / long   |
 | ldinfo_textorg  | 4; void *        / uint | 8; unsigned long long / void * |
 | ldinfo_textsize | 4; int                  | 8; long long          / long   |
 | ldinfo_dataorg  | 4; void *        / uint | 8; unsigned long long / void * |
 | ldinfo_datasize | 4; int                  | 8; long long          / long   |
 | ldinfo_filename | variable                | variable                       |
 ==============================================================================

> Anyway, I skimmed it and it looked good to me.

Thanks! I made the following adjustments to my patch making "next"
unsigned...

| diff --git a/gdb/rs6000-aix-tdep.c b/gdb/rs6000-aix-tdep.c
| index 450abae..9956eb3 100644
| --- a/gdb/rs6000-aix-tdep.c
| +++ b/gdb/rs6000-aix-tdep.c
| @@ -854,7 +854,7 @@ static const struct ld_info_desc ld_info64_desc =
|  
|  struct ld_info
|  {
| -  int next;
| +  ULONGEST next;
|    int fd;
|    CORE_ADDR textorg;
|    ULONGEST textsize;
| @@ -882,9 +882,9 @@ rs6000_aix_extract_ld_info (struct gdbarch *gdbarch,
|      = tdep->wordsize == 8 ? ld_info64_desc : ld_info32_desc;
|    struct ld_info info;
|  
| -  info.next = extract_signed_integer (ldi_buf + desc.ldinfo_next.offset,
| -				      desc.ldinfo_next.size,
| -				      byte_order);
| +  info.next = extract_unsigned_integer (ldi_buf + desc.ldinfo_next.offset,
| +					desc.ldinfo_next.size,
| +					byte_order);
|    info.fd = extract_signed_integer (ldi_buf + desc.ldinfo_fd.offset,
|  				    desc.ldinfo_fd.size,
|  				    byte_order);

... and checked the patch in.

For the record, attached is the patch I checked in.

Thanks for the quick review!
-- 
Joel