Re: [patch] bfd aix xty_ld symbols

[Nick Clifton <nickc at cambridge dot redhat dot com>]

Hi Tom,

> This patch fixes a bug with xty_ld symbols that do not immediately
> follow their xty_sd symbol.  
> 
> This bug may show up if you are using the .set pseudo op 
> 
> void foo() {}
> void bar() {}
> 
> .set .foo1, .foo

This appears to be a strange test file.  Is it a mix of C and
assembler or is it something else ?  Also if it is a test case then
why not include it in the patch so that the bug will be continue to be
fixed in future releases ?


> RCS file: /cvs/cvsfiles/devo/bfd/xcofflink.c,v
> *** 1706,1711 ****

You appear to be quoting a patch based against the Red Hat internal
sources and not the sourceware sources.  This is a bad idea.  In fact
you appear to have an out of date set of sources entirely since the
line you are patching is at 1604 in the sources not 1706.

>   
> + 	    csect = section;
>   	    value = sym.n_value - csect->vma;
>   	  }

I would also appreciate a little more explanation of what this patch
is doing and how it is fixing the problem.  For example in the code
you are patching there is a comment saying:

	case XTY_LD:
	  /* This is a label definition.  The x_scnlen field is the
             symbol index of the csect.  I believe that this must
             always follow the appropriate XTY_SD symbol, so I will
             insist on it.  */

So are you changing the behaviour of the linker, and if so, then why
not change the comment as well ?


> 2001-08-21  Tom Rix <trix@redhat.com> 
> 
> 	* xcofflink.c (xcoff_link_add_symbols):  Fix XTY_LD symbol that 
> 	does not follow a XTY_SD.

Not approved (yet).

Cheers
        Nick