Re: PATCH: Linker failed to handle symbols with mismatched types

Re: PATCH: Linker failed to handle symbols with mismatched types

[Alan Modra]

On Tue, Feb 28, 2006 at 01:48:22PM -0800, H. J. Lu wrote:
> 	PR ld/2404
> 	* elflink.c (_bfd_elf_merge_symbol): Skip the default indirect
> 	symbol from the dynamic definition with the default version if
> 	its type and the type of exiting regular definition mismatch.

I think the concept is reasonable, but I'd like to hear Ian's opinion
before giving an OK.  Also, I'm wondering if you should check for type
mismatches earlier, to avoid errors with tls syms.

-- 
Alan Modra
IBM OzLabs - Linux Technology Centre

Re: PATCH: Linker failed to handle symbols with mismatched types

[H. J. Lu]

On Thu, Mar 02, 2006 at 12:43:17PM +1030, Alan Modra wrote:
> On Tue, Feb 28, 2006 at 01:48:22PM -0800, H. J. Lu wrote:
> > 	PR ld/2404
> > 	* elflink.c (_bfd_elf_merge_symbol): Skip the default indirect
> > 	symbol from the dynamic definition with the default version if
> > 	its type and the type of exiting regular definition mismatch.
> 
> I think the concept is reasonable, but I'd like to hear Ian's opinion
> before giving an OK.  Also, I'm wondering if you should check for type
> mismatches earlier, to avoid errors with tls syms.

Here is the updated patch.


H.J.
----
2006-03-02  H.J. Lu  <hongjiu.lu@intel.com>

	PR ld/2404
	* elflink.c (_bfd_elf_merge_symbol): Skip the default indirect
	symbol from the dynamic definition with the default version if
	its type and the type of exiting regular definition mismatch.

--- bfd/elflink.c.mis	2006-03-02 06:33:15.000000000 -0800
+++ bfd/elflink.c	2006-03-02 06:35:31.000000000 -0800
@@ -888,6 +888,22 @@ _bfd_elf_merge_symbol (bfd *abfd,
 	    && h->root.type != bfd_link_hash_undefweak
 	    && h->root.type != bfd_link_hash_common);
 
+  /* When we try to create a default indirect symbol from the dynamic
+     definition with the default version, we skip it if its type and
+     the type of exiting regular definition mismatch.  */
+  if (pold_alignment == NULL
+      && newdyn
+      && newdef
+      && !olddyn
+      && (olddef || h->root.type == bfd_link_hash_common)
+      && ELF_ST_TYPE (sym->st_info) != h->type
+      && ELF_ST_TYPE (sym->st_info) != STT_NOTYPE
+      && h->type != STT_NOTYPE)
+    {
+      *skip = TRUE;
+      return TRUE;
+    }
+
   /* Check TLS symbol.  We don't check undefined symbol introduced by
      "ld -u".  */
   if ((ELF_ST_TYPE (sym->st_info) == STT_TLS || h->type == STT_TLS)

Re: PATCH: Linker failed to handle symbols with mismatched types

[H. J. Lu]

On Thu, Mar 02, 2006 at 06:39:10AM -0800, H. J. Lu wrote:
> On Thu, Mar 02, 2006 at 12:43:17PM +1030, Alan Modra wrote:
> > On Tue, Feb 28, 2006 at 01:48:22PM -0800, H. J. Lu wrote:
> > > 	PR ld/2404
> > > 	* elflink.c (_bfd_elf_merge_symbol): Skip the default indirect
> > > 	symbol from the dynamic definition with the default version if
> > > 	its type and the type of exiting regular definition mismatch.
> > 
> > I think the concept is reasonable, but I'd like to hear Ian's opinion
> > before giving an OK.  Also, I'm wondering if you should check for type
> > mismatches earlier, to avoid errors with tls syms.
> 
> Here is the updated patch.
> 

Can someone please take a look at

http://sourceware.org/ml/binutils/2006-03/msg00018.html

Thanks.


H.J.

Re: PATCH: Linker failed to handle symbols with mismatched types

[H. J. Lu]

On Thu, Mar 09, 2006 at 06:30:12PM -0800, H. J. Lu wrote:
> On Thu, Mar 02, 2006 at 06:39:10AM -0800, H. J. Lu wrote:
> > On Thu, Mar 02, 2006 at 12:43:17PM +1030, Alan Modra wrote:
> > > On Tue, Feb 28, 2006 at 01:48:22PM -0800, H. J. Lu wrote:
> > > > 	PR ld/2404
> > > > 	* elflink.c (_bfd_elf_merge_symbol): Skip the default indirect
> > > > 	symbol from the dynamic definition with the default version if
> > > > 	its type and the type of exiting regular definition mismatch.
> > > 
> > > I think the concept is reasonable, but I'd like to hear Ian's opinion
> > > before giving an OK.  Also, I'm wondering if you should check for type
> > > mismatches earlier, to avoid errors with tls syms.
> > 
> > Here is the updated patch.
> > 
> 
> Can someone please take a look at
> 
> http://sourceware.org/ml/binutils/2006-03/msg00018.html

Here is the updated patch. We only skip the default indirect symbol if
the exiting regular definition won't be dynamic.


H.J.
----
2006-03-09  H.J. Lu  <hongjiu.lu@intel.com>

	PR ld/2404
	* elflink.c (_bfd_elf_merge_symbol): Skip the default indirect
	symbol from the dynamic definition with the default version if
	its type and the type of exiting regular definition mismatch.

--- bfd/elflink.c.mis	2006-03-09 09:43:10.000000000 -0800
+++ bfd/elflink.c	2006-03-09 19:41:16.000000000 -0800
@@ -888,6 +888,26 @@ _bfd_elf_merge_symbol (bfd *abfd,
 	    && h->root.type != bfd_link_hash_undefweak
 	    && h->root.type != bfd_link_hash_common);
 
+  /* When we try to create a default indirect symbol from the dynamic
+     definition with the default version, we skip it if its type and
+     the type of exiting regular definition mismatch.  We only do it
+     if the exiting regular definition won't be dynamic.  */
+  if (pold_alignment == NULL
+      && !info->shared
+      && !info->export_dynamic
+      && !h->ref_dynamic
+      && newdyn
+      && newdef
+      && !olddyn
+      && (olddef || h->root.type == bfd_link_hash_common)
+      && ELF_ST_TYPE (sym->st_info) != h->type
+      && ELF_ST_TYPE (sym->st_info) != STT_NOTYPE
+      && h->type != STT_NOTYPE)
+    {
+      *skip = TRUE;
+      return TRUE;
+    }
+
   /* Check TLS symbol.  We don't check undefined symbol introduced by
      "ld -u".  */
   if ((ELF_ST_TYPE (sym->st_info) == STT_TLS || h->type == STT_TLS)

Re: PATCH: Linker failed to handle symbols with mismatched types

[Nick Clifton]

Hi H. J.

> 2006-03-09  H.J. Lu  <hongjiu.lu@intel.com>
> 
> 	PR ld/2404
> 	* elflink.c (_bfd_elf_merge_symbol): Skip the default indirect
> 	symbol from the dynamic definition with the default version if
> 	its type and the type of exiting regular definition mismatch.

Approved - please apply, but ...

I think that you have a spelling typo in the ChangeLog and the comment 
in the patch: exiting -> existing.

Cheers
   Nick