Dlltool fix - do not munge padding entries in .reloc section

Dlltool fix - do not munge padding entries in .reloc section

[Nick Clifton]

Hi Guys

  Does anyone have any objections to my applying the following small
  patch ?  It prevents flush_page() from munging padding entries in
  the .reloc section so that they come out as 0x0.  (Without the patch
  they come out as 0xf000, which is an illegal entry).

Cheers
	Nick


1999-11-18  Nick Clifton  <nickc@cygnus.com>

	* dlltool.c (flush_page): Do not munge padding entries in the
	.reloc section.

Index: binutils/dlltool.c
===================================================================
RCS file: /cvs/binutils/binutils/binutils/dlltool.c,v
retrieving revision 1.16
diff -p -r1.16 dlltool.c
*** dlltool.c	1999/11/03 18:16:52	1.16
--- dlltool.c	1999/11/18 18:32:52
*************** flush_page (f, need, page_addr, on_page)
*** 1574,1581 ****
  	   ASM_C);
    
    for (i = 0; i < on_page; i++)
!     fprintf (f, "\t%s\t0x%lx\n", ASM_SHORT,
! 	     ((need[i] - page_addr) | 0x3000) & 0xffff);
  
    /* And padding */
    if (on_page & 1)
--- 1574,1587 ----
  	   ASM_C);
    
    for (i = 0; i < on_page; i++)
!     {
!       long needed = need[i];
!       
!       if (needed)
! 	needed = ((needed - page_addr) | 0x3000) & 0xffff;
!   
!       fprintf (f, "\t%s\t0x%lx\n", ASM_SHORT, needed);
!     }
  
    /* And padding */
    if (on_page & 1)

Re: Dlltool fix - do not munge padding entries in .reloc section

[DJ Delorie]

Looks OK to me.  I know it's not likely, but what if a reloc really
does happen at address zero?

> 	* dlltool.c (flush_page): Do not munge padding entries in the
> 	.reloc section.

Re: Dlltool

[Lynn Winebarger]

  Speaking of DLL tools, is there any hope of full DLL support being
integrated into BFD itself?  Specifically, is there any hope that the
symbols a DLL defines will get recognized by BFD (they weren't last time
I tried an objdump -tT on a DLL, though dumping the sections containing
the symbols did seem to recognize the contents as symbols)?  I mean, any
hope short of me putting it in myself (unlikely any time soon, I'm
afraid).
   One problem I did notice is that exported and global symbols are
considered equivalent by BFD, whereas they aren't necessarily in DLL's (I
think - I could probably stand a closer reading of the MS doc describing
the PE format).

Lynn

Re: Dlltool

[Nick Clifton]

Hi Lynn,

:   Speaking of DLL tools, is there any hope of full DLL support being
: integrated into BFD itself?  Specifically, is there any hope that the
: symbols a DLL defines will get recognized by BFD (they weren't last time
: I tried an objdump -tT on a DLL, though dumping the sections containing
: the symbols did seem to recognize the contents as symbols)?

Do you have an example of this not working ?

I ask because it seems to be working for me.  At least I do get a
symbol dump with 'objdump -t foo.dll' using the current sources in
sourceware.

Cheers
	Nick