rpms/gdb/F-13 gdb-false-warning-gnu.liblist.patch, NONE, 1.1 gdb-stale-sectindex.patch, NONE, 1.1 gdb.spec, 1.410, 1.411

Jan Kratochvil jkratoch at fedoraproject.org
Sun Feb 28 23:47:01 UTC 2010


Author: jkratoch

Update of /cvs/pkgs/rpms/gdb/F-13
In directory cvs1.fedora.phx.redhat.com:/tmp/cvs-serv12344

Modified Files:
	gdb.spec 
Added Files:
	gdb-false-warning-gnu.liblist.patch gdb-stale-sectindex.patch 
Log Message:
* Sun Feb 28 2010 Jan Kratochvil <jan.kratochvil at redhat.com> - 7.0.50.20100203-17.fc13
- Fix false warning: section .gnu.liblist not found in ...
- Fix crash on stale addrinfo->sectindex (more sensitive due to the PIE patch).


gdb-false-warning-gnu.liblist.patch:
 symfile.c |   15 ++++++++++++---
 1 file changed, 12 insertions(+), 3 deletions(-)

--- NEW FILE gdb-false-warning-gnu.liblist.patch ---
http://sourceware.org/ml/gdb-patches/2010-02/msg00713.html
Subject: Re: [patch] Fix false warning: section .gnu.liblist not found in ...  [rediffed]

Hi,

only a technical rediff.

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

On Sat, 13 Feb 2010 23:49:29 +0100, Jan Kratochvil wrote:
Hi,

post-7.0 GDB started to print many false warnings on prelink-ed system:

$ ./gdb -nx -ex r echo
...
Reading symbols from /bin/echo...Reading symbols from /usr/lib/debug/bin/echo.debug...
warning: section .gnu.liblist not found in /usr/lib/debug/bin/echo.debug
warning: section .gnu.conflict not found in /usr/lib/debug/bin/echo.debug
done.
done.
Starting program: /bin/echo
warning: section .gnu.liblist not found in /usr/lib/debug/bin/echo.debug
warning: section .gnu.conflict not found in /usr/lib/debug/bin/echo.debug
warning: section .gnu.liblist not found in /usr/lib/debug/bin/echo.debug
warning: section .gnu.conflict not found in /usr/lib/debug/bin/echo.debug
warning: section .gnu.liblist not found in /usr/lib/debug/bin/echo.debug
warning: section .gnu.conflict not found in /usr/lib/debug/bin/echo.debug

These sections are from prelink and they are not present in the *.debug files.

It is formally a regression by:

commit 3bfec189bb0fa1a2a44f1645dd68a9572e7a841c

2010-01-07  Tristan Gingold  <gingold at adacore.com>

	* symfile.c (build_section_addr_info_from_objfile): New function.
	(symbol_file_add_separate): Don't use offsets from objfile but
	built an addr info.


But I think would be unavoidable even for the unification of PIC+PIE handling.

Found no usable section flag differences for .gnu.liblist vs. for example
.text so used just the section name for the exception.

.gnu.liblist/.gnu.conflict related dumps:

executable:
Section Headers:
  [Nr] Name              Type            Address          Off    Size   ES Flg Lk Inf Al
  [ 6] .gnu.liblist      GNU_LIBLIST     00000000004006b8 0006b8 000028 14   A 28   0  4
  [ 7] .gnu.conflict     RELA            00000000004006e0 0006e0 0001c8 18   A  5   0  8
Program Headers:
  Type           Offset   VirtAddr           PhysAddr           FileSiz  MemSiz   Flg Align
  LOAD           0x000000 0x0000000000400000 0x0000000000400000 0x005254 0x005254 R E 0x200000
Dynamic section at offset 0x52c8 contains 24 entries:
  Tag        Type                         Name/Value
 0x000000006ffffef9 (GNU_LIBLIST)        0x4006b8
 0x000000006ffffdf7 (GNU_LIBLISTSZ)      40 (bytes)
 0x000000006ffffef8 (GNU_CONFLICT)       0x4006e0
 0x000000006ffffdf6 (GNU_CONFLICTSZ)     456 (bytes)

library:
Section Headers:
  [Nr] Name              Type            Address          Off    Size   ES Flg Lk Inf Al
  [74] .gnu.liblist      GNU_LIBLIST     0000000000000000 249940 000014 14     75   0  4


No regressions on {x86_64,x86_64-m32,i686}-fedora12-linux-gnu.


Thanks,
Jan


2010-03-01  Jan Kratochvil  <jan.kratochvil at redhat.com>

	* symfile.c (addr_info_make_relative): New variable sect_name, use it.
	Do not warn on ".gnu.liblist" and ".gnu.conflict".

--- a/gdb/symfile.c
+++ b/gdb/symfile.c
@@ -597,7 +597,8 @@ addr_info_make_relative (struct section_addr_info *addrs, bfd *abfd)
 
   for (i = 0; i < addrs->num_sections && addrs->other[i].name; i++)
     {
-      asection *sect = bfd_get_section_by_name (abfd, addrs->other[i].name);
+      const char *sect_name = addrs->other[i].name;
+      asection *sect = bfd_get_section_by_name (abfd, sect_name);
 
       if (sect)
 	{
@@ -614,8 +615,16 @@ addr_info_make_relative (struct section_addr_info *addrs, bfd *abfd)
 	}
       else
 	{
-	  warning (_("section %s not found in %s"), addrs->other[i].name,
-		   bfd_get_filename (abfd));
+	  /* These two sections are intentionally loaded into memory from
+	     the DYNAMIC segment and so they have both SEC_ALLOC and SEC_LOAD
+	     set in the main executable (not in the library files).  They
+	     are not present in the separate debug info file, though.  */
+
+	  if (!(strcmp (sect_name, ".gnu.liblist") == 0
+		|| strcmp (sect_name, ".gnu.conflict") == 0))
+	    warning (_("section %s not found in %s"), sect_name,
+		     bfd_get_filename (abfd));
+
 	  addrs->other[i].addr = 0;
 
 	  /* SECTINDEX is invalid if ADDR is zero.  */


gdb-stale-sectindex.patch:
 symfile.c |   30 +++++++++++++++++-------------
 symfile.h |    2 ++
 2 files changed, 19 insertions(+), 13 deletions(-)

--- NEW FILE gdb-stale-sectindex.patch ---
http://sourceware.org/ml/gdb-patches/2010-02/msg00477.html
Subject: Re: [patch] Fix crash on stale addrinfo->sectindex

On Fri, 19 Feb 2010 04:01:05 +0100, Tom Tromey wrote:
> >>>>> "Jan" == Jan Kratochvil <jan.kratochvil at redhat.com> writes:
> I think you ought to move this comment as well.
> It appears to just be hanging at the end of a block after the patch.

Yes; forgot.

> This is ok with that change.  Thanks.

Checked-in.

On Fri, 19 Feb 2010 04:10:30 +0100, Tom Tromey wrote:
> Oh, by the way, I think this is reasonable for 7.1.  It is
> straightforward and fixes a reported crash.  If Joel agrees, please put
> it on the branch.  Thanks.

OK for the branch?


Thanks,
Jan


http://sourceware.org/ml/gdb-cvs/2010-02/msg00162.html

### src/gdb/ChangeLog	2010/02/19 00:35:53	1.11380
### src/gdb/ChangeLog	2010/02/19 06:19:44	1.11381
## -1,3 +1,10 @@
+2010-02-19  Jan Kratochvil  <jan.kratochvil at redhat.com>
+
+	* symfile.c (addr_info_make_relative): Extend comment.  Move SECT to
+	a more inner block.  Initialize ADDR by LOWER_OFFSET only if it was
+	found by bfd_get_section_by_name.
+	* symfile.h (struct section_addr_info) <sectindex>: New comment.
+
 2010-02-19  Joel Brobecker  <brobecker at adacore.com>
 
 	* NEWS: Add new "[...] since 7.1" section.  Rename the "[...] since
--- src/gdb/symfile.c	2010/02/18 19:17:00	1.272
+++ src/gdb/symfile.c	2010/02/19 06:19:45	1.273
@@ -562,13 +562,13 @@
 }
 
 /* Relativize absolute addresses in ADDRS into offsets based on ABFD.  Fill-in
-   also SECTINDEXes there.  */
+   also SECTINDEXes specific to ABFD there.  This function can be used to
+   rebase ADDRS to start referencing different BFD than before.  */
 
 void
 addr_info_make_relative (struct section_addr_info *addrs, bfd *abfd)
 {
   asection *lower_sect;
-  asection *sect;
   CORE_ADDR lower_offset;
   int i;
 
@@ -597,25 +597,29 @@
 
   for (i = 0; i < addrs->num_sections && addrs->other[i].name; i++)
     {
-      if (addrs->other[i].addr != 0)
+      asection *sect = bfd_get_section_by_name (abfd, addrs->other[i].name);
+
+      if (sect)
 	{
-	  sect = bfd_get_section_by_name (abfd, addrs->other[i].name);
-	  if (sect)
+	  /* This is the index used by BFD. */
+	  addrs->other[i].sectindex = sect->index;
+
+	  if (addrs->other[i].addr != 0)
 	    {
 	      addrs->other[i].addr -= bfd_section_vma (abfd, sect);
 	      lower_offset = addrs->other[i].addr;
-	      /* This is the index used by BFD. */
-	      addrs->other[i].sectindex = sect->index;
 	    }
 	  else
-	    {
-	      warning (_("section %s not found in %s"), addrs->other[i].name,
-		       bfd_get_filename (abfd));
-	      addrs->other[i].addr = 0;
-	    }
+	    addrs->other[i].addr = lower_offset;
 	}
       else
-	addrs->other[i].addr = lower_offset;
+	{
+	  warning (_("section %s not found in %s"), addrs->other[i].name,
+		   bfd_get_filename (abfd));
+	  addrs->other[i].addr = 0;
+
+	  /* SECTINDEX is invalid if ADDR is zero.  */
+	}
     }
 }
 
--- src/gdb/symfile.h	2010/02/03 14:13:16	1.64
+++ src/gdb/symfile.h	2010/02/19 06:19:45	1.65
@@ -80,6 +80,8 @@
   {
     CORE_ADDR addr;
     char *name;
+
+    /* SECTINDEX must be valid for associated BFD if ADDR is not zero.  */
     int sectindex;
   } other[1];
 };



Index: gdb.spec
===================================================================
RCS file: /cvs/pkgs/rpms/gdb/F-13/gdb.spec,v
retrieving revision 1.410
retrieving revision 1.411
diff -u -p -r1.410 -r1.411
--- gdb.spec	26 Feb 2010 08:56:34 -0000	1.410
+++ gdb.spec	28 Feb 2010 23:46:59 -0000	1.411
@@ -36,7 +36,7 @@ Version: 7.0.50.20100203
 
 # The release always contains a leading reserved number, start it at 1.
 # `upstream' is not a part of `name' to stay fully rpm dependencies compatible for the testing.
-Release: 16%{?_with_upstream:.upstream}%{dist}
+Release: 17%{?_with_upstream:.upstream}%{dist}
 
 License: GPLv3+
 Group: Development/Debuggers
@@ -430,9 +430,15 @@ Patch412: gdb-unused-revert.patch
 # Fix i386+x86_64 rwatch+awatch before run, regression against 6.8 (BZ 541866).
 Patch417: gdb-bz541866-rwatch-before-run.patch
 
+# Fix crash on stale addrinfo->sectindex (more sensitive due to the PIE patch).
+Patch421: gdb-stale-sectindex.patch
+
 # Remove false gdb_assert on $sp underflow.
 Patch422: gdb-infcall-sp-underflow.patch
 
+# Fix false warning: section .gnu.liblist not found in ...
+Patch425: gdb-false-warning-gnu.liblist.patch
+
 BuildRequires: ncurses-devel%{?_isa} texinfo gettext flex bison expat-devel%{?_isa}
 Requires: readline%{?_isa}
 BuildRequires: readline-devel%{?_isa}
@@ -683,7 +689,9 @@ rm -f gdb/jv-exp.c gdb/m2-exp.c gdb/objc
 %patch408 -p1
 %patch412 -p1
 %patch417 -p1
+%patch421 -p1
 %patch422 -p1
+%patch425 -p1
 
 %patch415 -p1
 %patch393 -p1
@@ -1016,6 +1024,10 @@ fi
 %endif
 
 %changelog
+* Sun Feb 28 2010 Jan Kratochvil <jan.kratochvil at redhat.com> - 7.0.50.20100203-17.fc13
+- Fix false warning: section .gnu.liblist not found in ...
+- Fix crash on stale addrinfo->sectindex (more sensitive due to the PIE patch).
+
 * Fri Feb 26 2010 Jan Kratochvil <jan.kratochvil at redhat.com> - 7.0.50.20100203-16.fc13
 - Fix ia64 part of the bt-clone-stop.exp fix.
 - Fix gdb.ada/* regressions (Keith Seitz).



More information about the scm-commits mailing list