[patch] Fix C++ no .eh_frame_hdr table will be created (PR 6893)

[patch] Fix C++ no .eh_frame_hdr table will be created (PR 6893)

[Jan Kratochvil]

[-- Attachment #1: Type: text/plain, Size: 2242 bytes --]

Hi,

original bugreport by Andrew Haley:

1. svn co http://llvm.org/svn/llvm-project/llvm/trunk llvm
2. cd llvm
3. ./configure \
--disable-static \
--enable-assertions \
--enable-debug-runtime \
--enable-jit \
--enable-optimized \
--enable-shared \
--enable-targets=host-only \
--with-pic \
--enable-pic \
--disable-binding
4. make VERBOSE=true

Actual results:/usr/bin/ld: error in
/home/aph/llvm/Release/lib/LLVMInterpreter.o(.eh_frame); no .eh_frame_hdr table will be created.
/usr/bin/ld: error in /home/aph/llvm/Release/lib/LLVMX86.o(.eh_frame); no .eh_frame_hdr table will be created.
/usr/bin/ld: error in /home/aph/llvm/Release/lib/LLVMExecutionEngine.o(.eh_frame); no .eh_frame_hdr table will be created.
/usr/bin/ld: error in /home/aph/llvm/Release/lib/LLVMJIT.o(.eh_frame); no .eh_frame_hdr table will be created.

------------------------------------------------------------------------------

These relocations
Relocation section '.rela.eh_frame' at offset 0x14b0 contains 2 entries:
  Offset          Info           Type           Sym. Value    Sym. Name +
Addend
000000000020  000600000002 R_X86_64_PC32     0000000000000000 .text._ZNK4llvm13Targe + 0

with its target section .text._ZNK4llvm13 which is SHT_GROUP
  [ 9] .text._ZNK4llvm13 PROGBITS         0000000000000000  000004c0
       000000000000004f  0000000000000000 AXG       0     0     16

get converted to:
000000000020  000000000000 R_X86_64_NONE                       0000000000000000

by this code:
../bfd/elf64-x86-64.c
      if (sec != NULL && elf_discarded_section (sec))
        {
          /* For relocs against symbols from removed linkonce sections,
             or sections discarded by a linker script, we just want the
             section contents zeroed.  Avoid any special processing.  */
          _bfd_clear_contents (howto, input_bfd, contents + rel->r_offset);
          rel->r_info = 0;
          rel->r_addend = 0;
          continue;
        }

ld will later error out on R_X86_64_NONE


The attached fix will properly discard those FDEs with discarded relocations
which can be verified in the testcase.

original bugreports:
https://bugzilla.redhat.com/show_bug.cgi?id=458950
https://bugzilla.redhat.com/show_bug.cgi?id=461675


Thanks,
Jan

[-- Attachment #2: llvm-binutils4.patch --]
[-- Type: text/plain, Size: 6882 bytes --]

bfd/
2008-09-16  Jan Kratochvil  <jan.kratochvil@redhat.com>

	PR 6893 - Do not consider FDEs for discarded sections as invalid.
	* elf-eh-frame.c (_bfd_elf_parse_eh_frame): New REQUIRE_CLEARED_RELOCS.
	Consider FDEs with cleared relocations as valid and ignorable.

ld/testsuite/
2008-09-16  Jan Kratochvil  <jan.kratochvil@redhat.com>

	* ld-elf/eh-group.exp, ld-elf/eh-group1.s, ld-elf/eh-group2.s: New test.

binutils/
2008-09-16  Jan Kratochvil  <jan.kratochvil@redhat.com>

	Suppress warnings on NONE relocations to discarded sections.
	* readelf.c (is_none_reloc): New function.
	(debug_apply_relocations): Ignore is_none_reloc() relocations.

--- bfd/elf-eh-frame.c	24 Aug 2008 21:43:00 -0000	1.72
+++ bfd/elf-eh-frame.c	16 Sep 2008 13:28:43 -0000
@@ -549,6 +549,16 @@ _bfd_elf_parse_eh_frame (bfd *abfd, stru
 	     < (bfd_size_type) ((buf) - ehbuf)))	\
     cookie->rel++
 
+#define REQUIRE_CLEARED_RELOCS(buf)			\
+  while (cookie->rel < cookie->relend			\
+	 && (cookie->rel->r_offset			\
+	     < (bfd_size_type) ((buf) - ehbuf)))	\
+    {							\
+      REQUIRE (cookie->rel->r_info == 0);		\
+      REQUIRE (cookie->rel->r_addend == 0);		\
+      cookie->rel++;					\
+    }
+
 #define GET_RELOC(buf)					\
   ((cookie->rel < cookie->relend			\
     && (cookie->rel->r_offset				\
@@ -766,9 +776,14 @@ _bfd_elf_parse_eh_frame (bfd *abfd, stru
 
 	  /* Chain together the FDEs for each section.  */
 	  rsec = _bfd_elf_gc_mark_rsec (info, sec, gc_mark_hook, cookie);
-	  REQUIRE (rsec && rsec->owner == abfd);
-	  this_inf->u.fde.next_for_section = elf_fde_list (rsec);
-	  elf_fde_list (rsec) = this_inf;
+	  /* RSEC will be NULL if FDE was cleared out as it was belonging to
+	     a discarded SHT_GROUP.  */
+	  if (rsec)
+	    {
+	      REQUIRE (rsec->owner == abfd);
+	      this_inf->u.fde.next_for_section = elf_fde_list (rsec);
+	      elf_fde_list (rsec) = this_inf;
+	    }
 
 	  /* Skip the initial location and address range.  */
 	  start = buf;
@@ -801,7 +816,17 @@ _bfd_elf_parse_eh_frame (bfd *abfd, stru
 	  insns = buf;
 
 	  buf = last_fde + 4 + hdr_length;
-	  SKIP_RELOCS (buf);
+
+	  /* Cleared FDE?  The instructions will not be cleared but verify all
+	     the relocation entries for them are cleared.  */
+	  if (rsec == NULL)
+	    {
+	      REQUIRE_CLEARED_RELOCS (buf);
+	    }
+	  else
+	    {
+	      SKIP_RELOCS (buf);
+	    }
 	}
 
       /* Try to interpret the CFA instructions and find the first
--- binutils/readelf.c	5 Sep 2008 14:49:05 -0000	1.425
+++ binutils/readelf.c	16 Sep 2008 13:28:50 -0000
@@ -8284,6 +8284,53 @@ is_16bit_abs_reloc (unsigned int reloc_t
     }
 }
 
+/* Returns TRUE iff RELOC_TYPE is a NONE relocation used for discarded
+   relocation entries (possibly formerly used for SHT_GROUP sections).  */
+
+static bfd_boolean
+is_none_reloc (unsigned int reloc_type)
+{
+  switch (elf_header.e_machine)
+    {
+    case EM_68K:
+      return reloc_type == 0; /* R_68K_NONE.  */
+    case EM_386:
+      return reloc_type == 0; /* R_386_NONE.  */
+    case EM_SPARC32PLUS:
+    case EM_SPARCV9:
+    case EM_SPARC:
+      return reloc_type == 0; /* R_SPARC_NONE.  */
+    case EM_MIPS:
+      return reloc_type == 0; /* R_MIPS_NONE.  */
+    case EM_PARISC:
+      return reloc_type == 0; /* R_PARISC_NONE.  */
+    case EM_ALPHA:
+      return reloc_type == 0; /* R_ALPHA_NONE.  */
+    case EM_PPC:
+      return reloc_type == 0; /* R_PPC_NONE.  */
+    case EM_PPC64:
+      return reloc_type == 0; /* R_PPC64_NONE.  */
+    case EM_ARM:
+      return reloc_type == 0; /* R_ARM_NONE.  */
+    case EM_IA_64:
+      return reloc_type == 0; /* R_IA64_NONE.  */
+    case EM_SH:
+      return reloc_type == 0; /* R_SH_NONE.  */
+    case EM_S390_OLD:
+    case EM_S390:
+      return reloc_type == 0; /* R_390_NONE.  */
+    case EM_CRIS:
+      return reloc_type == 0; /* R_CRIS_NONE.  */
+    case EM_X86_64:
+      return reloc_type == 0; /* R_X86_64_NONE.  */
+    case EM_MN10300:
+      return reloc_type == 0; /* R_MN10300_NONE.  */
+    case EM_M32R:
+      return reloc_type == 0; /* R_M32R_NONE.  */
+    }
+  return FALSE;
+}
+
 /* Uncompresses a section that was compressed using zlib, in place.
  * This is a copy of bfd_uncompress_section_contents, in bfd/compress.c  */
 
@@ -8419,6 +8466,9 @@ debug_apply_relocations (void *file,
 
 	  reloc_type = get_reloc_type (rp->r_info);
 
+	  if (is_none_reloc (reloc_type))
+	    continue;
+
 	  if (is_32bit_abs_reloc (reloc_type)
 	      || is_32bit_pcrel_reloc (reloc_type))
 	    reloc_size = 4;
--- /dev/null	1 Jan 1970 00:00:00 -0000
+++ ld/testsuite/ld-elf/eh-group.exp	16 Sep 2008 13:28:54 -0000
@@ -0,0 +1,51 @@
+# Expect script for .eh_frame entries to a removed section.
+#   Copyright 2008  Free Software Foundation, Inc.
+#
+# This file is part of the GNU Binutils.
+#
+# This program is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 3 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program; if not, write to the Free Software
+# Foundation, Inc., 51 Franklin Street - Fifth Floor, Boston,
+# MA 02110-1301, USA.
+#
+
+#
+# Written by Jan Kratochvil (jan.kratochvil@redhat.com)
+#
+# .eh_frame with relocations to a removed (group) section did result to:
+# error in tmpdir/eh-group.o(.eh_frame); no .eh_frame_hdr table will be created.
+# The purpose of this test is to merge two .o files with -r and then link this
+# merged file (containing a discarded R_X86_64_NONE relocation) to the final
+# executable trying to create .eh_frame_hdr.  It needs a separate .exp file due
+# to the requirement of two `ld' runs.
+
+# Exclude non-ELF targets.
+
+if ![is_elf_format] {
+    return
+}
+
+set build_tests_ld {
+  {"Build eh-group1.o"
+   "-r" ""
+   {eh-group1.s eh-group2.s} {} "eh-group.o"}
+}
+
+run_ld_link_tests $build_tests_ld
+
+set testname "Link eh-group.o to eh-group"
+if [ld_simple_link $ld "tmpdir/eh-group" "-e _start tmpdir/eh-group.o"] {
+    pass $testname
+} else {
+    fail $testname
+}
--- /dev/null	1 Jan 1970 00:00:00 -0000
+++ ld/testsuite/ld-elf/eh-group1.s	16 Sep 2008 13:28:54 -0000
@@ -0,0 +1,6 @@
+	.section	sect, "axG", @progbits, sectgroup, comdat
+	.global	_start
+_start:
+	.cfi_startproc
+	.skip 16
+	.cfi_endproc
--- /dev/null	1 Jan 1970 00:00:00 -0000
+++ ld/testsuite/ld-elf/eh-group2.s	16 Sep 2008 13:28:54 -0000
@@ -0,0 +1,4 @@
+	.section	sect, "axG", @progbits, sectgroup, comdat
+	.cfi_startproc
+	.skip 16
+	.cfi_endproc

Re: [patch] Fix C++ no .eh_frame_hdr table will be created (PR 6893)

[Alan Modra]

On Tue, Sep 16, 2008 at 04:08:07PM +0200, Jan Kratochvil wrote:
> The attached fix will properly discard those FDEs with discarded relocations
> which can be verified in the testcase.

Please apply.  2.19 branch too.

-- 
Alan Modra
Australia Development Lab, IBM

Re: [patch] Fix C++ no .eh_frame_hdr table will be created (PR 6893)

[Andreas Schwab]

Jan Kratochvil <jan.kratochvil@redhat.com> writes:

> +# Exclude non-ELF targets.
> +
> +if ![is_elf_format] {
> +    return
> +}
> +
> +set build_tests_ld {
> +  {"Build eh-group1.o"
> +   "-r" ""
> +   {eh-group1.s eh-group2.s} {} "eh-group.o"}
> +}
> +
> +run_ld_link_tests $build_tests_ld
> +
> +set testname "Link eh-group.o to eh-group"
> +if [ld_simple_link $ld "tmpdir/eh-group" "-e _start tmpdir/eh-group.o"] {
> +    pass $testname
> +} else {
> +    fail $testname
> +}
> --- /dev/null	1 Jan 1970 00:00:00 -0000
> +++ ld/testsuite/ld-elf/eh-group1.s	16 Sep 2008 13:28:54 -0000
> @@ -0,0 +1,6 @@
> +	.section	sect, "axG", @progbits, sectgroup, comdat
> +	.global	_start
> +_start:
> +	.cfi_startproc
> +	.skip 16
> +	.cfi_endproc
> --- /dev/null	1 Jan 1970 00:00:00 -0000
> +++ ld/testsuite/ld-elf/eh-group2.s	16 Sep 2008 13:28:54 -0000
> @@ -0,0 +1,4 @@
> +	.section	sect, "axG", @progbits, sectgroup, comdat
> +	.cfi_startproc
> +	.skip 16
> +	.cfi_endproc

This fails on ia64, because .cfi_* does not exist.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux Products GmbH, Maxfeldstraße 5, 90409 Nürnberg, Germany
PGP key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: [patch] Fix C++ no .eh_frame_hdr table will be created (PR 6893)

[Andreas Schwab]

Jan Kratochvil <jan.kratochvil@redhat.com> writes:

> --- /dev/null	1 Jan 1970 00:00:00 -0000
> +++ ld/testsuite/ld-elf/eh-group1.s	16 Sep 2008 13:28:54 -0000
> @@ -0,0 +1,6 @@
> +	.section	sect, "axG", @progbits, sectgroup, comdat
> +	.global	_start
> +_start:
> +	.cfi_startproc
> +	.skip 16
> +	.cfi_endproc
> --- /dev/null	1 Jan 1970 00:00:00 -0000
> +++ ld/testsuite/ld-elf/eh-group2.s	16 Sep 2008 13:28:54 -0000
> @@ -0,0 +1,4 @@
> +	.section	sect, "axG", @progbits, sectgroup, comdat
> +	.cfi_startproc
> +	.skip 16
> +	.cfi_endproc

This will also fail to assemble on arm because @ is a comment character.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux Products GmbH, Maxfeldstraße 5, 90409 Nürnberg, Germany
PGP key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: [patch] Fix C++ no .eh_frame_hdr table will be created (PR 6893)

[Jan Kratochvil]

On Wed, 17 Sep 2008 17:41:07 +0200, Andreas Schwab wrote:
> Jan Kratochvil <jan.kratochvil@redhat.com> writes:
> > +# Exclude non-ELF targets.
> > +
> > +if ![is_elf_format] {
> > +    return
> > +}
> 
> This fails on ia64, because .cfi_* does not exist.

On Wed, 17 Sep 2008 17:50:10 +0200, Andreas Schwab wrote:
> Jan Kratochvil <jan.kratochvil@redhat.com> writes:
> > +	.section	sect, "axG", @progbits, sectgroup, comdat
> 
> This will also fail to assemble on arm because @ is a comment character.


I see now the other CFI-using testcases use an explicit list of targets.
Tested those listed below.

OK to apply?


Thanks,
Jan


2008-09-17  Jan Kratochvil  <jan.kratochvil@redhat.com>

	* ld-elf/eh-group.exp: Execute only on tested targets supporting CFI.

--- ld/testsuite/ld-elf/eh-group.exp	17 Sep 2008 07:50:29 -0000	1.1
+++ ld/testsuite/ld-elf/eh-group.exp	17 Sep 2008 17:28:43 -0000
@@ -29,9 +29,13 @@
 # executable trying to create .eh_frame_hdr.  It needs a separate .exp file due
 # to the requirement of two `ld' runs.
 
-# Exclude non-ELF targets.
+# Exclude non-ELF targets and targets not supporting CFI.
 
-if ![is_elf_format] {
+if { ![is_elf_format] \
+     || (![istarget i?86-*-linux*] \
+	 && ![istarget powerpc*-*-linux*] \
+	 && ![istarget s390*-*-linux*] \
+	 && ![istarget x86_64-*-linux*]) } {
     return
 }
 

Re: [patch] Fix C++ no .eh_frame_hdr table will be created (PR 6893)

[Andreas Schwab]

Jan Kratochvil <jan.kratochvil@redhat.com> writes:

> -if ![is_elf_format] {
> +if { ![is_elf_format] \
> +     || (![istarget i?86-*-linux*] \
> +	 && ![istarget powerpc*-*-linux*] \
> +	 && ![istarget s390*-*-linux*] \
> +	 && ![istarget x86_64-*-linux*]) } {

There are more targets supporting cfi (all those that define
TARGET_USE_CFIPOP).

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux Products GmbH, Maxfeldstraße 5, 90409 Nürnberg, Germany
PGP key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: [patch] Fix C++ no .eh_frame_hdr table will be created (PR 6893)

[Nick Clifton]

Hi Jan,

>> -if ![is_elf_format] {
>> +if { ![is_elf_format] \
>> +     || (![istarget i?86-*-linux*] \
>> +	 && ![istarget powerpc*-*-linux*] \
>> +	 && ![istarget s390*-*-linux*] \
>> +	 && ![istarget x86_64-*-linux*]) } {
> 
> There are more targets supporting cfi (all those that define
> TARGET_USE_CFIPOP).

If this test is needed for more than just one file then we ought to move 
it into ld/testsuite/lib/ld-lib.exp so that bug fixes and additions can 
be added in just one place.

Cheers
   Nick

Re: [patch] Fix C++ no .eh_frame_hdr table will be created (PR 6893)

[Jan Kratochvil]

[-- Attachment #1: Type: text/plain, Size: 845 bytes --]

Hi Nick,

On Thu, 18 Sep 2008 09:15:56 +0200, Nick Clifton wrote:
>>> -if ![is_elf_format] {
>>> +if { ![is_elf_format] \
>>> +     || (![istarget i?86-*-linux*] \
>>> +	 && ![istarget powerpc*-*-linux*] \
>>> +	 && ![istarget s390*-*-linux*] \
>>> +	 && ![istarget x86_64-*-linux*]) } {
>>
>> There are more targets supporting cfi (all those that define
>> TARGET_USE_CFIPOP).
>
> If this test is needed for more than just one file then we ought to move  
> it into ld/testsuite/lib/ld-lib.exp so that bug fixes and additions can  
> be added in just one place.

Originally I just copied the style I saw there.  Here is the general framework
for CFI tests.
Tested on: {alpha,arm,m68k,i386,ppc,sh,s390,sparc,hppa,mips}-linux

2.19 should IMO get only the ia64-fixing patch:
	http://sourceware.org/ml/binutils/2008-09/msg00138.html


Thanks,
Jan

[-- Attachment #2: binutils-cfi.patch --]
[-- Type: text/plain, Size: 5854 bytes --]

2008-09-18  Jan Kratochvil  <jan.kratochvil@redhat.com>

	Provide virtual target "cfi" for targets supporting CFI.
	* ld-elf/eh-frame-hdr.d: Replace target and xfail statements by single
	`target: cfi'.
	* ld-elf/eh-group.exp: Call check_as_cfi instead of is_elf_format.
	* ld-elf/eh-group1.s, elf/eh-group2.s: Use more compatible section
	flags prefix '%'.
	* ld-elf/eh5.d: Replace target statement by `target: cfi'.  Relax the
	`Code alignment factor' matching.
	* lib/ld-lib.exp: Rename istarget as istarget_ld.
	(istarget, check_as_cfi): New procedure.
	(run_dump_test): New comment for the virtual target `cfi'.

--- ld/testsuite/ld-elf/eh-frame-hdr.d	4 Dec 2006 08:57:09 -0000	1.1
+++ ld/testsuite/ld-elf/eh-frame-hdr.d	18 Sep 2008 20:40:59 -0000
@@ -1,17 +1,7 @@
 #source: eh-frame-hdr.s
 #ld: -e _start --eh-frame-hdr
 #objdump: -hw
-#target: alpha*-*-*
-#target: arm*-*-*
-#target: i?86-*-*
-#target: m68k-*-*
-#target: mips*-*-*
-#target: powerpc*-*-*
-#target: s390*-*-*
-#target: sh*-*-*
-#xfail: sh*l*-*-*
-#target: sparc*-*-*
-#target: x86_64-*-*
+#target: cfi
 #...
   [0-9] .eh_frame_hdr 0*[12][048c] .*
 #pass
--- ld/testsuite/ld-elf/eh-group.exp	17 Sep 2008 07:50:29 -0000	1.1
+++ ld/testsuite/ld-elf/eh-group.exp	18 Sep 2008 20:40:59 -0000
@@ -29,9 +29,9 @@
 # executable trying to create .eh_frame_hdr.  It needs a separate .exp file due
 # to the requirement of two `ld' runs.
 
-# Exclude non-ELF targets.
+# Exclude non-CFI (such as ia64) targets.
 
-if ![is_elf_format] {
+if {![check_as_cfi]} {
     return
 }
 
--- ld/testsuite/ld-elf/eh-group1.s	17 Sep 2008 07:50:29 -0000	1.1
+++ ld/testsuite/ld-elf/eh-group1.s	18 Sep 2008 20:40:59 -0000
@@ -1,4 +1,4 @@
-	.section	sect, "axG", @progbits, sectgroup, comdat
+	.section	sect, "axG", %progbits, sectgroup, comdat
 	.global	_start
 _start:
 	.cfi_startproc
--- ld/testsuite/ld-elf/eh-group2.s	17 Sep 2008 07:50:29 -0000	1.1
+++ ld/testsuite/ld-elf/eh-group2.s	18 Sep 2008 20:40:59 -0000
@@ -1,4 +1,4 @@
-	.section	sect, "axG", @progbits, sectgroup, comdat
+	.section	sect, "axG", %progbits, sectgroup, comdat
 	.cfi_startproc
 	.skip 16
 	.cfi_endproc
--- ld/testsuite/ld-elf/eh5.d	28 Jan 2008 15:15:32 -0000	1.2
+++ ld/testsuite/ld-elf/eh5.d	18 Sep 2008 20:40:59 -0000
@@ -3,14 +3,14 @@
 #source: eh5b.s
 #ld:
 #readelf: -wf
-#target: x86_64-*-* i?86-*-*
+#target: cfi
 
 The section .eh_frame contains:
 
 00000000 0000001[04] 00000000 CIE
   Version:               1
   Augmentation:          "zR"
-  Code alignment factor: 1
+  Code alignment factor: .*
   Data alignment factor: .*
   Return address column: .*
   Augmentation data:     1b
@@ -29,7 +29,7 @@ The section .eh_frame contains:
 000000(2c|30) 00000014 00000000 CIE
   Version:               1
   Augmentation:          "zPR"
-  Code alignment factor: 1
+  Code alignment factor: .*
   Data alignment factor: .*
   Return address column: .*
   Augmentation data:     03 .. .. .. .. 1b
@@ -53,7 +53,7 @@ The section .eh_frame contains:
 0000007[48] 0000001[8c] 00000000 CIE
   Version:               1
   Augmentation:          "zPLR"
-  Code alignment factor: 1
+  Code alignment factor: .*
   Data alignment factor: .*
   Return address column: .*
   Augmentation data:     03 .. .. .. .. 0c 1b
@@ -74,7 +74,7 @@ The section .eh_frame contains:
 000000b[08] 0000001[04] 00000000 CIE
   Version:               1
   Augmentation:          "zR"
-  Code alignment factor: 1
+  Code alignment factor: .*
   Data alignment factor: .*
   Return address column: .*
   Augmentation data:     1b
@@ -89,7 +89,7 @@ The section .eh_frame contains:
 000000[de]8 00000014 00000000 CIE
   Version:               1
   Augmentation:          "zPR"
-  Code alignment factor: 1
+  Code alignment factor: .*
   Data alignment factor: .*
   Return address column: .*
   Augmentation data:     03 .. .. .. .. 1b
@@ -111,7 +111,7 @@ The section .eh_frame contains:
 000001(1c|30) 0000001[8c] 00000000 CIE
   Version:               1
   Augmentation:          "zPLR"
-  Code alignment factor: 1
+  Code alignment factor: .*
   Data alignment factor: .*
   Return address column: .*
   Augmentation data:     03 .. .. .. .. 0c 1b
--- ld/testsuite/lib/ld-lib.exp	18 Feb 2008 11:04:09 -0000	1.54
+++ ld/testsuite/lib/ld-lib.exp	18 Sep 2008 20:41:01 -0000
@@ -598,7 +598,8 @@ proc simple_diff { file_1 file_2 } {
 #
 #   target: TARGET
 #       Only run the test for TARGET.  This may occur more than once; the
-#       target being tested must match at least one.
+#       target being tested must match at least one.  You may provide target
+#       name "cfi" for any target supporting the CFI statements.
 #
 #   notarget: TARGET
 #       Do not run the test for TARGET.  This may occur more than once;
@@ -1569,3 +1570,39 @@ proc check_gc_sections_available { } {
     }
     return $gc_sections_available_saved
 }
+
+# Check if the assembler supports CFI statements.
+
+proc check_as_cfi { } {
+    global check_as_cfi_result
+    global as
+    if [info exists check_as_cfi_result] {
+	return $check_as_cfi_result
+    }
+    set as_file "tmpdir/check_as_cfi.s"
+    set as_fh [open $as_file w 0666]
+    puts $as_fh "# Generated file. DO NOT EDIT"
+    puts $as_fh "\t.cfi_startproc"
+    puts $as_fh "\t.cfi_endproc"
+    close $as_fh
+    remote_download host $as_file
+    verbose -log "Checking CFI support:"
+    rename "perror" "check_as_cfi_perror"
+    proc perror { args } { }
+    set success [ld_assemble $as $as_file "/dev/null"]
+    rename "perror" ""
+    rename "check_as_cfi_perror" "perror"
+    #remote_file host delete $as_file
+    set check_as_cfi_result $success
+    return $success
+}
+
+# Provide virtual target "cfi" for targets supporting CFI.
+
+rename "istarget" "istarget_ld"
+proc istarget { target } {
+    if {$target == "cfi"} {
+	return [check_as_cfi]
+    }
+    return [istarget_ld $target]
+}

Re: [patch] Fix C++ no .eh_frame_hdr table will be created (PR 6893)

[Alan Modra]

On Thu, Sep 18, 2008 at 11:18:35PM +0200, Jan Kratochvil wrote:
> 	Provide virtual target "cfi" for targets supporting CFI.
> 	* ld-elf/eh-frame-hdr.d: Replace target and xfail statements by single
> 	`target: cfi'.
> 	* ld-elf/eh-group.exp: Call check_as_cfi instead of is_elf_format.
> 	* ld-elf/eh-group1.s, elf/eh-group2.s: Use more compatible section
> 	flags prefix '%'.
> 	* ld-elf/eh5.d: Replace target statement by `target: cfi'.  Relax the
> 	`Code alignment factor' matching.
> 	* lib/ld-lib.exp: Rename istarget as istarget_ld.
> 	(istarget, check_as_cfi): New procedure.
> 	(run_dump_test): New comment for the virtual target `cfi'.

OK.

-- 
Alan Modra
Australia Development Lab, IBM