rpms/gdb/F-13 gdb-bz614604-bt-cfi-without-die.patch, NONE, 1.1 gdb-bz614659-prelink-dynbss.patch, NONE, 1.1 gdb.spec, 1.446, 1.447
Jan Kratochvil
jkratoch at fedoraproject.org
Tue Jul 20 17:21:23 UTC 2010
Author: jkratoch
Update of /cvs/pkgs/rpms/gdb/F-13
In directory cvs01.phx2.fedoraproject.org:/tmp/cvs-serv31375
Modified Files:
gdb.spec
Added Files:
gdb-bz614604-bt-cfi-without-die.patch
gdb-bz614659-prelink-dynbss.patch
Log Message:
* Tue Jul 20 2010 Jan Kratochvil <jan.kratochvil at redhat.com> - 7.1-29.fc13
- Fix prelinked executables with sepdebug and copy relocations (BZ 614659).
- [delayed-symfile] Fix a backtrace regression on CFIs without DIE (BZ 614604).
gdb-bz614604-bt-cfi-without-die.patch:
dwarf2-frame.c | 8 +--
testsuite/gdb.base/cfi-without-die-caller.c | 28 +++++++++++
testsuite/gdb.base/cfi-without-die-main.c | 32 +++++++++++++
testsuite/gdb.base/cfi-without-die.exp | 68 +++++++++++++++++++++++++++-
4 files changed, 130 insertions(+), 6 deletions(-)
--- NEW FILE gdb-bz614604-bt-cfi-without-die.patch ---
http://sourceware.org/ml/archer/2010-q3/msg00028.html
Subject: [delayed-symfile] [commit] Fix a regression on CFI without DIE [Re:
On Wed, 25 Feb 2009 00:14:29 +0100, Jan Kratochvil wrote:
> commit 6a37c2b9962258ecf9299cc34a650e64a06acaa5
>
> There was a regression on gdb.base/savedregs.exp.
>
> quick_addrmap/require_partial_symbols should be used even for the unwind debug
> info checking as its load has been also delayed by this branch.
[...]
> --- a/gdb/dwarf2-frame.c
> +++ b/gdb/dwarf2-frame.c
[...]
> @@ -1499,6 +1500,14 @@ dwarf2_frame_find_fde (CORE_ADDR *pc)
> struct dwarf2_fde *fde;
> CORE_ADDR offset;
>
> + if (objfile->quick_addrmap)
> + {
> + if (!addrmap_find (objfile->quick_addrmap, *pc))
> + continue;
> + }
> + /* FIXME: Read-in only .debug_frame/.eh_frame without .debug_info? */
> + require_partial_symbols (objfile);
> +
but this has caused a different regression (as discussed in the confcall).
QUICK_ADDRMAP is built only from .debug_aranges. But we can have existing
built .debug_aranges for CUs in OBJFILE but still some CUs do not need to have
DWARF at all while they can feature CFIs (.eh_frame or .debug_frame).
It has been described by Daniel Jacobowitz at:
Re: [2/4] RFC: check psymtabs_addrmap before reading FDEs
http://sourceware.org/ml/gdb-patches/2010-07/msg00012.html
Sorry for this regression by me (in that fix of a different regression).
Fixed it the "slow way" as this branch is now obsoleted by .gdb-index.
No regressions on {x86_64,x86_64-m32,i686}-fedora13-linux-gnu.
Checked-in.
Thanks,
Jan
eb8df8566acc1ed963e3e9b77c13b9c2c3db03fb
Test CFI is parsed even for range (function) not described by any DIE.
https://bugzilla.redhat.com/show_bug.cgi?id=614028
gdb/
* dwarf2-frame.c (dwarf2_frame_find_fde): Remove the
OBJFILE->QUICK_ADDRMAP check. New comment why.
gdb/testsuite/
* gdb.base/cfi-without-die.exp, gdb.base/cfi-without-die-main.c,
gdb.base/cfi-without-die-caller.c: New files.
---
gdb/dwarf2-frame.c | 8 +--
gdb/testsuite/gdb.base/cfi-without-die-caller.c | 28 ++++++++++
gdb/testsuite/gdb.base/cfi-without-die-main.c | 32 +++++++++++
gdb/testsuite/gdb.base/cfi-without-die.exp | 67 +++++++++++++++++++++++
4 files changed, 130 insertions(+), 5 deletions(-)
create mode 100644 gdb/testsuite/gdb.base/cfi-without-die-caller.c
create mode 100644 gdb/testsuite/gdb.base/cfi-without-die-main.c
create mode 100644 gdb/testsuite/gdb.base/cfi-without-die.exp
diff --git a/gdb/dwarf2-frame.c b/gdb/dwarf2-frame.c
index 5915249..1dc2754 100644
--- a/gdb/dwarf2-frame.c
+++ b/gdb/dwarf2-frame.c
@@ -1583,11 +1583,9 @@ dwarf2_frame_find_fde (CORE_ADDR *pc)
CORE_ADDR offset;
CORE_ADDR seek_pc;
- if (objfile->quick_addrmap)
- {
- if (!addrmap_find (objfile->quick_addrmap, *pc))
- continue;
- }
+ /* OBJFILE->QUICK_ADDRMAP contains offsets only for DIEs. It does not
+ contain ranges of CFIs. */
+
/* FIXME: Read-in only .debug_frame/.eh_frame without .debug_info? */
require_partial_symbols (objfile);
diff --git a/gdb/testsuite/gdb.base/cfi-without-die-caller.c b/gdb/testsuite/gdb.base/cfi-without-die-caller.c
new file mode 100644
index 0000000..afdfd53
--- /dev/null
+++ b/gdb/testsuite/gdb.base/cfi-without-die-caller.c
@@ -0,0 +1,28 @@
+/* This testcase is part of GDB, the GNU debugger.
+
+ Copyright 2005, 2007, 2008, 2009, 2010 Free Software Foundation, Inc.
+
+ 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, see <http://www.gnu.org/licenses/>. */
+
+typedef int (*callback_t) (void);
+
+int
+caller (callback_t callback)
+{
+ /* Ensure some frame content to push away the return address. */
+ volatile const long one = 1;
+
+ /* Modify the return value to prevent any tail-call optimization. */
+ return (*callback) () - one;
+}
diff --git a/gdb/testsuite/gdb.base/cfi-without-die-main.c b/gdb/testsuite/gdb.base/cfi-without-die-main.c
new file mode 100644
index 0000000..8451c4b
--- /dev/null
+++ b/gdb/testsuite/gdb.base/cfi-without-die-main.c
@@ -0,0 +1,32 @@
+/* This testcase is part of GDB, the GNU debugger.
+
+ Copyright 2005, 2007, 2008, 2009, 2010 Free Software Foundation, Inc.
+
+ 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, see <http://www.gnu.org/licenses/>. */
+
+typedef int (*callback_t) (void);
+
+extern int caller (callback_t callback);
+
+int
+callback (void)
+{
+ return 1;
+}
+
+int
+main (void)
+{
+ return caller (callback);
+}
diff --git a/gdb/testsuite/gdb.base/cfi-without-die.exp b/gdb/testsuite/gdb.base/cfi-without-die.exp
new file mode 100644
index 0000000..db6d248
--- /dev/null
+++ b/gdb/testsuite/gdb.base/cfi-without-die.exp
@@ -0,0 +1,67 @@
+# Copyright 2010 Free Software Foundation, Inc.
+
+# 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, see <http://www.gnu.org/licenses/>.
+
+# Test CFI is parsed even for range (function) not described by any DIE.
+
+set testfile cfi-without-die
+set srcmainfile ${testfile}-main.c
+set srccallerfile ${testfile}-caller.c
+set executable ${testfile}
+set objmainfile ${objdir}/${subdir}/${testfile}-main.o
+set objcallerfile ${objdir}/${subdir}/${testfile}-caller.o
+set binfile ${objdir}/${subdir}/${executable}
+
+if { [gdb_compile "${srcdir}/${subdir}/${srccallerfile}" ${objcallerfile} \
+ object [list {additional_flags=-fomit-frame-pointer -fno-unwind-tables -fno-asynchronous-unwind-tables}]] != ""
+ || [gdb_compile "${srcdir}/${subdir}/${srcmainfile}" ${objmainfile} object {debug}] != ""
+ || [gdb_compile "${objmainfile} ${objcallerfile}" ${binfile} executable {}] != "" } {
+ untested ${testfile}.exp
+ return -1
+}
+
+clean_restart $executable
+
+if ![runto callback] then {
+ fail "verify unwinding: Can't run to callback"
+ return 0
+}
+set test "verify unwinding breaks without CFI"
+gdb_test_multiple "bt" $test {
+ -re " in main .*\r\n$gdb_prompt $" {
+ fail $test
+ }
+ -re "\r\n$gdb_prompt $" {
+ pass $test
+ }
+}
+
+if { [gdb_compile "${srcdir}/${subdir}/${srccallerfile}" ${objcallerfile} \
+ object [list {additional_flags=-fomit-frame-pointer -funwind-tables -fasynchronous-unwind-tables}]] != ""
+ || [gdb_compile "${srcdir}/${subdir}/${srcmainfile}" ${objmainfile} object {debug}] != ""
+ || [gdb_compile "${objmainfile} ${objcallerfile}" ${binfile} executable {}] != "" } {
+ untested ${testfile}.exp
+ return -1
+}
+
+clean_restart $executable
+
+if ![runto callback] then {
+ fail "test CFI without DIEs: Can't run to callback"
+ return 0
+}
+# #0 callback () at ...
+# #1 0x00000000004004e9 in caller ()
+# #2 0x00000000004004cd in main () at ...
+gdb_test "bt" "#0 +callback \[^\r\n\]+\r\n#1 \[^\r\n\]+ in caller \[^\r\n\]+\r\n#2 \[^\r\n\]+ in main \[^\r\n\]+" "verify unwindin works for CFI without DIEs"
--
1.7.1.1
gdb-bz614659-prelink-dynbss.patch:
symfile.c | 42 ++++++++++++++++++++++++++++++++++-----
testsuite/gdb.base/prelink-lib.c | 2 +
testsuite/gdb.base/prelink.c | 6 +++++
testsuite/gdb.base/prelink.exp | 11 +++++++++-
4 files changed, 55 insertions(+), 6 deletions(-)
--- NEW FILE gdb-bz614659-prelink-dynbss.patch ---
http://sourceware.org/ml/gdb-patches/2010-07/msg00237.html
Subject: [patch] Fix regression on prelinked executables
Hi,
there is a regression since gdb-7.0 for a combination of:
* prelinked
* main executable
* using separate debug info
* using copy relocations
It is since a patch for both PIE and (AFAIK) OSX support:
[commit] syms_from_objfile: Relativize also MAINLINE
http://sourceware.org/ml/gdb-patches/2010-01/msg00080.html
which started to use problematic addr_info_make_relative even for main
executables. prelink<->gdb discussion at:
https://bugzilla.redhat.com/show_bug.cgi?id=614659
Currently in the unfortunately executables GDB has invalid displcement for
symbols in .bss:
int bssvar, *bssvarp = &bssvar;
(gdb) p &bssvar
$1 = (int *) 0x600b54
(gdb) p bssvarp
$2 = (int *) 0x600b50
<abstract-higher-point-of-view>
addr_info_make_relative could just simply subtract entry point address and
provide single CORE_ADDR objfile->offset (instead of the current
section_offsets array with offsets specific for each section). Linux systems
use always single offset for the whole objfile. AFAIK these per-section
offsets are there for some embedded targets. Curiously GDB already uses at
many places
baseaddr = ANOFFSET (objfile->section_offsets, SECT_OFF_TEXT (objfile));
instead of using offset for the appropriate section at that place and nobody
complains.
</abstract-higher-point-of-view>
No regressions on {x86_64,x86_64-m32,i686}-fedora13-linux-gnu.
Proposing for the gdb-7.2 branch. I had problems fixing up my crashing X.
Thanks,
Jan
gdb/
2010-07-15 Jan Kratochvil <jan.kratochvil at redhat.com>
* symfile.c (addr_section_name): New function.
(addrs_section_compar): Use it.
(addr_info_make_relative): Use it. Move variable sect_name into a more
inner block. Make ".dynbss" and ".sdynbss" checks more strict.
gdb/testsuite/
2010-07-15 Jan Kratochvil <jan.kratochvil at redhat.com>
* gdb.base/prelink-lib.c (copyreloc): New initialized variable.
* gdb.base/prelink.c (copyreloc, bssvar, bssvarp): New variables.
(main): Use copyreloc.
* gdb.base/prelink.exp (split debug of executable)
(.dynbss vs. .bss address shift): New tests.
Index: gdb-7.1/gdb/symfile.c
===================================================================
--- gdb-7.1.orig/gdb/symfile.c 2010-07-19 23:11:10.000000000 +0200
+++ gdb-7.1/gdb/symfile.c 2010-07-20 18:34:50.000000000 +0200
@@ -557,6 +557,23 @@ relative_addr_info_to_section_offsets (s
}
}
+/* Transform section name S for a name comparison. prelink can split section
+ `.bss' into two sections `.dynbss' and `.bss' (in this order). Similarly
+ prelink can split `.sbss' into `.sdynbss' and `.sbss'. Use virtual address
+ of the new `.dynbss' (`.sdynbss') section as the adjacent new `.bss'
+ (`.sbss') section has invalid (increased) virtual address. */
+
+static const char *
+addr_section_name (const char *s)
+{
+ if (strcmp (s, ".dynbss") == 0)
+ return ".bss";
+ if (strcmp (s, ".sdynbss") == 0)
+ return ".sbss";
+
+ return s;
+}
+
/* Relativize absolute addresses in ADDRS into offsets based on ABFD. Fill-in
also SECTINDEXes specific to ABFD there. This function can be used to
rebase ADDRS to start referencing different BFD than before. */
@@ -607,8 +624,17 @@ addr_info_make_relative (struct section_
if (sect && strcmp (sect_name, bfd_get_section_name (abfd, sect)) != 0)
sect = NULL;
- if (sect == NULL)
- sect = bfd_get_section_by_name (abfd, sect_name);
+ /* Prevent the search by name if `.bss' has the address already set from
+ `.dynbss'. */
+ if (sect == NULL
+ && !(0
+ || (strcmp (sect_name, ".bss") == 0
+ && i > 0
+ && strcmp (addrs->other[i - 1].name, ".dynbss") == 0)
+ || (strcmp (sect_name, ".sbss") == 0
+ && i > 0
+ && strcmp (addrs->other[i - 1].name, ".sdynbss") == 0)))
+ sect = bfd_get_section_by_name (abfd, addr_section_name (sect_name));
if (sect)
{
/* This is the index used by BFD. */
@@ -634,12 +660,18 @@ addr_info_make_relative (struct section_
a warning. Shared libraries contain just the section
".gnu.liblist" but it is not marked as loadable there. There is
no other way to identify them than by their name as the sections
- created by prelink have no special flags. */
+ created by prelink have no special flags.
+
+ For the sections `.bss' and `.sbss' see addr_section_name. */
if (!(strcmp (sect_name, ".gnu.liblist") == 0
|| strcmp (sect_name, ".gnu.conflict") == 0
- || strcmp (sect_name, ".dynbss") == 0
- || strcmp (sect_name, ".sdynbss") == 0))
+ || (strcmp (sect_name, ".bss") == 0
+ && i > 0
+ && strcmp (addrs->other[i - 1].name, ".dynbss") == 0)
+ || (strcmp (sect_name, ".sbss") == 0
+ && i > 0
+ && strcmp (addrs->other[i - 1].name, ".sdynbss") == 0)))
warning (_("section %s not found in %s"), sect_name,
bfd_get_filename (abfd));
Index: gdb-7.1/gdb/testsuite/gdb.base/prelink-lib.c
===================================================================
--- gdb-7.1.orig/gdb/testsuite/gdb.base/prelink-lib.c 2010-01-01 08:32:01.000000000 +0100
+++ gdb-7.1/gdb/testsuite/gdb.base/prelink-lib.c 2010-07-19 23:11:56.000000000 +0200
@@ -16,6 +16,8 @@
along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
+int copyreloc = 1;
+
int
g (void (*p)(void))
{
Index: gdb-7.1/gdb/testsuite/gdb.base/prelink.c
===================================================================
--- gdb-7.1.orig/gdb/testsuite/gdb.base/prelink.c 2010-01-01 08:32:01.000000000 +0100
+++ gdb-7.1/gdb/testsuite/gdb.base/prelink.c 2010-07-19 23:11:56.000000000 +0200
@@ -18,6 +18,11 @@
#include <stdio.h>
+extern int copyreloc;
+
+/* Test GDB itself finds `&bssvar' right. */
+static int bssvar, *bssvarp = &bssvar;
+
extern void (*h (void)) (void (*)(void));
int
@@ -25,5 +30,6 @@ main (void)
{
void (*f) (void (*)(void)) = h ();
printf ("%p\n", f);
+ printf ("%d\n", copyreloc);
f (0);
}
Index: gdb-7.1/gdb/testsuite/gdb.base/prelink.exp
===================================================================
--- gdb-7.1.orig/gdb/testsuite/gdb.base/prelink.exp 2010-07-19 23:11:12.000000000 +0200
+++ gdb-7.1/gdb/testsuite/gdb.base/prelink.exp 2010-07-20 00:06:18.000000000 +0200
@@ -84,6 +84,13 @@ if { [gdb_compile "${srcdir}/${subdir}/$
return -1;
}
+set test "split debug of executable"
+if [gdb_gnu_strip_debug $binfile] {
+ fail $test
+} else {
+ pass $test
+}
+
set found 0
set coredir "${objdir}/${subdir}/coredir.[getpid]"
file mkdir $coredir
@@ -118,7 +125,7 @@ if {[catch "system \"/usr/sbin/prelink -
untested "${testfile}.so was not prelinked, maybe system libraries are not prelinked?"
return 0
}
-catch "system \"/usr/sbin/prelink -qNR --no-exec-shield ${libfile}\""
+catch "system \"/usr/sbin/prelink -qNR --no-exec-shield ${libfile} ${binfile}\""
# Start with a fresh gdb
@@ -131,3 +138,5 @@ gdb_load ${binfile}
gdb_test "set verbose on"
gdb_test "core-file $objdir/$subdir/prelink.core" {Using PIC \(Position Independent Code\) prelink displacement.*} "prelink"
+
+gdb_test "p &bssvar == bssvarp" " = 1" ".dynbss vs. .bss address shift"
Index: gdb.spec
===================================================================
RCS file: /cvs/pkgs/rpms/gdb/F-13/gdb.spec,v
retrieving revision 1.446
retrieving revision 1.447
diff -u -p -r1.446 -r1.447
--- gdb.spec 30 Jun 2010 14:11:31 -0000 1.446
+++ gdb.spec 20 Jul 2010 17:21:23 -0000 1.447
@@ -36,7 +36,7 @@ Version: 7.1
# 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: 28%{?_with_upstream:.upstream}%{dist}
+Release: 29%{?_with_upstream:.upstream}%{dist}
License: GPLv3+ and GPLv3+ with exceptions and GPLv2+ and GPLv2+ with exceptions and GPL+ and LGPLv2+ and GFDL and BSD and Public Domain
Group: Development/Debuggers
@@ -527,6 +527,12 @@ Patch485: gdb-bz562763-pretty-print-2d-v
Patch486: gdb-bz562763-pretty-print-2d-vectors.patch
Patch487: gdb-bz562763-pretty-print-2d-vectors-libstdcxx.patch
+# Fix prelinked executables with sepdebug and copy relocations (BZ 614659).
+Patch489: gdb-bz614659-prelink-dynbss.patch
+
+# [delayed-symfile] Fix a backtrace regression on CFIs without DIE (BZ 614604).
+Patch490: gdb-bz614604-bt-cfi-without-die.patch
+
BuildRequires: ncurses-devel%{?_isa} texinfo gettext flex bison expat-devel%{?_isa}
Requires: readline%{?_isa}
BuildRequires: readline-devel%{?_isa}
@@ -829,6 +835,8 @@ rm -f gdb/jv-exp.c gdb/m2-exp.c gdb/objc
%patch485 -p1
%patch486 -p1
%patch487 -p1
+%patch489 -p1
+%patch490 -p1
%patch415 -p1
%patch393 -p1
@@ -1161,6 +1169,10 @@ fi
%endif
%changelog
+* Tue Jul 20 2010 Jan Kratochvil <jan.kratochvil at redhat.com> - 7.1-29.fc13
+- Fix prelinked executables with sepdebug and copy relocations (BZ 614659).
+- [delayed-symfile] Fix a backtrace regression on CFIs without DIE (BZ 614604).
+
* Wed Jun 30 2010 Jan Kratochvil <jan.kratochvil at redhat.com> - 7.1-28.fc13
- Print 2D C++ vectors as matrices (BZ 562763, sourceware10659, Chris Moller).
More information about the scm-commits
mailing list