* Re: [PATCH] visibility fixes for s390*.
@ 2003-05-09 15:11 Martin Schwidefsky
0 siblings, 0 replies; 5+ messages in thread
From: Martin Schwidefsky @ 2003-05-09 15:11 UTC (permalink / raw)
To: Daniel Jacobowitz; +Cc: binutils
> Please check any patches you would like added to 2.14 on the actual
> branch. I think the first two patches from today won't be needed.
Aha. Good that I asked ;-) So the patches go only into CVS head for now.
blue skies,
Martin
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] visibility fixes for s390*.
2003-05-09 15:05 Martin Schwidefsky
@ 2003-05-09 15:08 ` Daniel Jacobowitz
0 siblings, 0 replies; 5+ messages in thread
From: Daniel Jacobowitz @ 2003-05-09 15:08 UTC (permalink / raw)
To: Martin Schwidefsky; +Cc: binutils
On Fri, May 09, 2003 at 05:04:43PM +0200, Martin Schwidefsky wrote:
>
> > Are you sure this is applicable to the branch? I believe that none of
> > the other changes which required platform-specific modifications went
> > to the branch.
>
> Well, I debugged the problem on 2.14.90.0.1. And the problem is there.
2.14.90.0.1 is a HEAD snapshot with HJ's local changes. It has no
relationship to the 2.14 branch at all; the snapshots from that branch
are still marked 2.13.91.
Please check any patches you would like added to 2.14 on the actual
branch. I think the first two patches from today won't be needed.
--
Daniel Jacobowitz
MontaVista Software Debian GNU/Linux Developer
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] visibility fixes for s390*.
@ 2003-05-09 15:05 Martin Schwidefsky
2003-05-09 15:08 ` Daniel Jacobowitz
0 siblings, 1 reply; 5+ messages in thread
From: Martin Schwidefsky @ 2003-05-09 15:05 UTC (permalink / raw)
To: Daniel Jacobowitz; +Cc: binutils
> Are you sure this is applicable to the branch? I believe that none of
> the other changes which required platform-specific modifications went
> to the branch.
Well, I debugged the problem on 2.14.90.0.1. And the problem is there.
blue skies,
Martin
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] visibility fixes for s390*.
2003-05-09 14:55 Martin Schwidefsky
@ 2003-05-09 15:01 ` Daniel Jacobowitz
0 siblings, 0 replies; 5+ messages in thread
From: Daniel Jacobowitz @ 2003-05-09 15:01 UTC (permalink / raw)
To: Martin Schwidefsky; +Cc: binutils
On Fri, May 09, 2003 at 04:54:46PM +0200, Martin Schwidefsky wrote:
> Hi,
> I incorporated the visibility fixes from i386 to the s390 backends. This
> fixes some of the ld test case failures for s390*. This is applicable for
> CVS head and the 2.14 branch. Ok to commit ?
Are you sure this is applicable to the branch? I believe that none of
the other changes which required platform-specific modifications went
to the branch.
--
Daniel Jacobowitz
MontaVista Software Debian GNU/Linux Developer
^ permalink raw reply [flat|nested] 5+ messages in thread
* [PATCH] visibility fixes for s390*.
@ 2003-05-09 14:55 Martin Schwidefsky
2003-05-09 15:01 ` Daniel Jacobowitz
0 siblings, 1 reply; 5+ messages in thread
From: Martin Schwidefsky @ 2003-05-09 14:55 UTC (permalink / raw)
To: binutils
Hi,
I incorporated the visibility fixes from i386 to the s390 backends. This
fixes some of the ld test case failures for s390*. This is applicable for
CVS head and the 2.14 branch. Ok to commit ?
blue skies,
Martin.
2003-05-09 Martin Schwidefsky <schwidefsky@de.ibm.com>
* elf32-s390.c (allocate_dynrelocs): For undef weak syms with
non-default visibility, a) don't make them dynamic, b) discard
space for dynamic relocs.
(elf_s390_relocate_section): Initialize the GOT entries and skip
R_390_{8,16,32}/R_390_PC{16,16DBL,32DBL,32} for weak undefined
symbols with non-default visibility.
* elf64-s390.c: Likewise.
diff -urN src/bfd/elf32-s390.c src-s390/bfd/elf32-s390.c
--- src/bfd/elf32-s390.c Thu May 8 21:40:46 2003
+++ src-s390/bfd/elf32-s390.c Fri May 9 15:09:37 2003
@@ -1651,7 +1651,9 @@
htab = elf_s390_hash_table (info);
if (htab->elf.dynamic_sections_created
- && h->plt.refcount > 0)
+ && h->plt.refcount > 0
+ && (ELF_ST_VISIBILITY (h->other) == STV_DEFAULT
+ || h->root.type != bfd_link_hash_undefweak))
{
/* Make sure this symbol is output as a dynamic symbol.
Undefined weak syms won't yet be marked as dynamic. */
@@ -1662,7 +1664,8 @@
return FALSE;
}
- if (WILL_CALL_FINISH_DYNAMIC_SYMBOL (1, info, h))
+ if (info->shared
+ || WILL_CALL_FINISH_DYNAMIC_SYMBOL (1, info, h))
{
asection *s = htab->splt;
@@ -1758,7 +1761,10 @@
htab->srelgot->_raw_size += sizeof (Elf32_External_Rela);
else if (tls_type == GOT_TLS_GD)
htab->srelgot->_raw_size += 2 * sizeof (Elf32_External_Rela);
- else if (WILL_CALL_FINISH_DYNAMIC_SYMBOL (dyn, info, h))
+ else if ((ELF_ST_VISIBILITY (h->other) == STV_DEFAULT
+ || h->root.type != bfd_link_hash_undefweak)
+ && (info->shared
+ || WILL_CALL_FINISH_DYNAMIC_SYMBOL (dyn, 0, h)))
htab->srelgot->_raw_size += sizeof (Elf32_External_Rela);
}
else
@@ -1792,6 +1798,12 @@
pp = &p->next;
}
}
+
+ /* Also discard relocs on undefined weak syms with non-default
+ visibility. */
+ if (ELF_ST_VISIBILITY (h->other) != STV_DEFAULT
+ && h->root.type == bfd_link_hash_undefweak)
+ eh->dyn_relocs = NULL;
}
else
{
@@ -2306,7 +2318,9 @@
&& (info->symbolic
|| h->dynindx == -1
|| (h->elf_link_hash_flags & ELF_LINK_FORCED_LOCAL))
- && (h->elf_link_hash_flags & ELF_LINK_HASH_DEF_REGULAR)))
+ && (h->elf_link_hash_flags & ELF_LINK_HASH_DEF_REGULAR))
+ || (ELF_ST_VISIBILITY (h->other)
+ && h->root.type == bfd_link_hash_undefweak))
{
/* This is actually a static link, or it is a
-Bsymbolic link and the symbol is defined
@@ -2471,6 +2485,9 @@
break;
if ((info->shared
+ && (h == NULL
+ || ELF_ST_VISIBILITY (h->other) == STV_DEFAULT
+ || h->root.type != bfd_link_hash_undefweak)
&& ((r_type != R_390_PC16
&& r_type != R_390_PC16DBL
&& r_type != R_390_PC32DBL
diff -urN src/bfd/elf64-s390.c src-s390/bfd/elf64-s390.c
--- src/bfd/elf64-s390.c Thu May 8 22:14:07 2003
+++ src-s390/bfd/elf64-s390.c Fri May 9 15:09:51 2003
@@ -1623,7 +1623,9 @@
htab = elf_s390_hash_table (info);
if (htab->elf.dynamic_sections_created
- && h->plt.refcount > 0)
+ && h->plt.refcount > 0
+ && (ELF_ST_VISIBILITY (h->other) == STV_DEFAULT
+ || h->root.type != bfd_link_hash_undefweak))
{
/* Make sure this symbol is output as a dynamic symbol.
Undefined weak syms won't yet be marked as dynamic. */
@@ -1634,7 +1636,8 @@
return FALSE;
}
- if (WILL_CALL_FINISH_DYNAMIC_SYMBOL (1, info, h))
+ if (info->shared
+ || WILL_CALL_FINISH_DYNAMIC_SYMBOL (1, info, h))
{
asection *s = htab->splt;
@@ -1730,7 +1733,10 @@
htab->srelgot->_raw_size += sizeof (Elf64_External_Rela);
else if (tls_type == GOT_TLS_GD)
htab->srelgot->_raw_size += 2 * sizeof (Elf64_External_Rela);
- else if (WILL_CALL_FINISH_DYNAMIC_SYMBOL (dyn, info, h))
+ else if ((ELF_ST_VISIBILITY (h->other) == STV_DEFAULT
+ || h->root.type != bfd_link_hash_undefweak)
+ && (info->shared
+ || WILL_CALL_FINISH_DYNAMIC_SYMBOL (dyn, 0, h)))
htab->srelgot->_raw_size += sizeof (Elf64_External_Rela);
}
else
@@ -1764,6 +1770,12 @@
pp = &p->next;
}
}
+
+ /* Also discard relocs on undefined weak syms with non-default
+ visibility. */
+ if (ELF_ST_VISIBILITY (h->other) != STV_DEFAULT
+ && h->root.type == bfd_link_hash_undefweak)
+ eh->dyn_relocs = NULL;
}
else
{
@@ -2279,7 +2291,9 @@
&& (info->symbolic
|| h->dynindx == -1
|| (h->elf_link_hash_flags & ELF_LINK_FORCED_LOCAL))
- && (h->elf_link_hash_flags & ELF_LINK_HASH_DEF_REGULAR)))
+ && (h->elf_link_hash_flags & ELF_LINK_HASH_DEF_REGULAR))
+ || (ELF_ST_VISIBILITY (h->other)
+ && h->root.type == bfd_link_hash_undefweak))
{
/* This is actually a static link, or it is a
-Bsymbolic link and the symbol is defined
@@ -2449,6 +2463,9 @@
break;
if ((info->shared
+ && (h == NULL
+ || ELF_ST_VISIBILITY (h->other) == STV_DEFAULT
+ || h->root.type != bfd_link_hash_undefweak)
&& ((r_type != R_390_PC16
&& r_type != R_390_PC16DBL
&& r_type != R_390_PC32
^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2003-05-09 15:11 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2003-05-09 15:11 [PATCH] visibility fixes for s390* Martin Schwidefsky
-- strict thread matches above, loose matches on Subject: below --
2003-05-09 15:05 Martin Schwidefsky
2003-05-09 15:08 ` Daniel Jacobowitz
2003-05-09 14:55 Martin Schwidefsky
2003-05-09 15:01 ` Daniel Jacobowitz
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).