This is the mail archive of the
binutils@sourceware.org
mailing list for the binutils project.
Re: PATCH: PR gas/11789: Bogus line debug info
H.J.,
On Mon, 5 Jul 2010, Alan Modra wrote:
> On Sun, Jul 04, 2010 at 11:26:01AM -0700, H.J. Lu wrote:
> > binutils/testsuite/
> >
> > 2009-10-27 H.J. Lu <hongjiu.lu@intel.com>
> >
> > PR gas/10531
> > * binutils-all/objdump.W: Remove bogus line debug info.
> >
> > gas/
> >
> > 2009-10-27 Jim Wilson <wilson@codesourcery.com>
> >
> > PR gas/10531
> > * dwarf2dbg.c (dwarf2_finish): Don't generate .debug_line section
> > if it isn't empty.
> >
> > gas/testsuite/
> >
> > 2010-07-04 H.J. Lu <hongjiu.lu@intel.com>
> >
> > * gas/i386/dw2-compress-1.d: Remove bogus line debug info.
> >
> > 2009-10-27 H.J. Lu <hongjiu.lu@intel.com>
> >
> > PR gas/10531
> > * gas/elf/dwarf2-1.d: New.
> > * gas/elf/dwarf2-1.s: Likewise.
> > * gas/elf/dwarf2-2.d: Likewise.
> > * gas/elf/dwarf2-2.s: Likewise.
> > * gas/elf/dwarf2-3.d: Likewise.
> > * gas/elf/dwarf2-3.s: Likewise.
> > * gas/i386/debug1.d: Likewise.
> > * gas/i386/debug1.s: Likewise.
> >
> > * gas/elf/elf.exp: Run dwarf2-1, dwarf2-2 and dwarf2-3.
> >
> > * gas/i386/i386.exp: Run debug1 for both 32bit and 64bit.
>
> OK.
This caused a regression on MIPS targets:
mips-sde-elf-as -o dump.o [...]/gas/testsuite/gas/elf/dwarf2-3.s
Executing on host: sh -c {mips-sde-elf-as -o dump.o [...]/gas/testsuite/gas/elf/dwarf2-3.s 2>&1} /dev/null gas.out (timeout = 300)
mips-sde-elf-readelf -wl dump.o
Executing on host: sh -c {mips-sde-elf-readelf -wl dump.o >dump.out 2>gas.stderr} /dev/null (timeout = 300)
readelf: Warning: Corrupt unit length (0) found in section .debug_info
readelf: Warning: Unable to load/parse the .debug_info section, so cannot interpret the .debug_line section.
FAIL: DWARF2 3
mips-gnu-linux is likewise affected.
A brief look at the contents of your .debug_info section reveals
non-standard usage of platform-specific GAS directives to emit DWARF-2
records. You should be using .2byte, .4byte, etc., that have been
invented specifically for DWARF records, rather than .short, .long, etc.,
whose behaviour (size, alignment) is target-dependent.
The fix below removes the regression, but please be more careful in the
future. Thanks.
2010-07-09 Maciej W. Rozycki <macro@codesourcery.com>
gas/testsuite/
* gas/elf/dwarf2-3.s: Replace .short and .long directives in
.debug_info section with .2byte and .4byte respectively.
OK to apply?
Maciej
binutils-fsf-2.20.51-20100709-dwarf2-fix-0.patch
--- gas/testsuite/gas/elf/dwarf2-3.s 2010-07-09 03:43:48.000000000 +0100
+++ gas/testsuite/gas/elf/dwarf2-3.s 2010-07-09 03:51:20.000000000 +0100
@@ -22,68 +22,68 @@ _evoke_link_warning_foo:
.text
.Letext0:
.section .debug_info
- .long 0x8a
- .short 0x2
- .long .Ldebug_abbrev0
+ .4byte 0x8a
+ .2byte 0x2
+ .4byte .Ldebug_abbrev0
.byte 0x4
.uleb128 0x1
- .long .Ldebug_line0
- .long .Letext0
- .long .Ltext0
- .long .LASF4
+ .4byte .Ldebug_line0
+ .4byte .Letext0
+ .4byte .Ltext0
+ .4byte .LASF4
.byte 0x1
- .long .LASF5
+ .4byte .LASF5
.uleb128 0x2
- .long 0x31
- .long 0x38
+ .4byte 0x31
+ .4byte 0x38
.uleb128 0x3
- .long 0x31
+ .4byte 0x31
.byte 0x1a
.byte 0x0
.uleb128 0x4
- .long .LASF0
+ .4byte .LASF0
.byte 0x4
.byte 0x7
.uleb128 0x5
- .long 0x3d
+ .4byte 0x3d
.uleb128 0x4
- .long .LASF1
+ .4byte .LASF1
.byte 0x1
.byte 0x6
.uleb128 0x6
- .long .LASF2
+ .4byte .LASF2
.byte 0x1
.byte 0x3
- .long 0x55
+ .4byte 0x55
.byte 0x5
.byte 0x3
- .long _evoke_link_warning_foo
+ .4byte _evoke_link_warning_foo
.uleb128 0x5
- .long 0x21
+ .4byte 0x21
.uleb128 0x2
- .long 0x6a
- .long 0x6c
+ .4byte 0x6a
+ .4byte 0x6c
.uleb128 0x3
- .long 0x31
+ .4byte 0x31
.byte 0x0
.byte 0x0
.uleb128 0x7
.byte 0x1
.uleb128 0x5
- .long 0x71
+ .4byte 0x71
.uleb128 0x8
.byte 0x4
- .long 0x6a
+ .4byte 0x6a
.uleb128 0x6
- .long .LASF3
+ .4byte .LASF3
.byte 0x1
.byte 0x9
- .long 0x88
+ .4byte 0x88
.byte 0x5
.byte 0x3
- .long init_array
+ .4byte init_array
.uleb128 0x5
- .long 0x5a
+ .4byte 0x5a
.byte 0x0
.section .debug_abbrev
.uleb128 0x1