From: "Tomar, Sourabh Singh via Gdb-patches" <gdb-patches@sourceware.org>
To: Simon Marchi <simon.marchi@polymtl.ca>,
Keith Seitz <keiths@redhat.com>,
"gdb-patches@sourceware.org" <gdb-patches@sourceware.org>
Subject: RE: [PATCH] Fix macro info lookup for binaries containing DWARFv5 line table
Date: Fri, 14 May 2021 19:20:11 +0000 [thread overview]
Message-ID: <DM4PR12MB5295FE3DA86D7D26DD10F4D89D509@DM4PR12MB5295.namprd12.prod.outlook.com> (raw)
In-Reply-To: <5ee2be96-e428-2c4d-2be1-8147922c85eb@polymtl.ca>
[AMD Public Use]
Hi Simon,
>It's not clear how you produced this output (which compiler and which command). What's the version of that section? Because it looks like a DWARF 4 or before section. GCC 10 produces a DWARF 3 >.debug_line section, even when using -gdwarf-5.
>In DWARF 4 and before, a dir_index of 0 referred to the compilation directory, even if there wasn't an explicit dir[0] value. See section 6.2.4, bullet 12, of DWARF4.pdf. So you would look up the >DW_AT_comp attribute of the compilation unit, append "macro.c", and that would give you the location of that file.
Apologies for sharing sparse details of the tools/commands.
Details: Distribution : 18.04.1-Ubuntu
GCC: rebased and **built today** i.e gcc (GCC) 12.0.0 20210514 (experimental)
Build config:
./configure --disable-bootstrap --build=x86_64-linux-gnu --host=x86_64-linux-gnu --target=x86_64-linux-gnu CFLAGS=-O3 CXXFLAGS=-O3 --enable-languages=c,c++,fortran --prefix=/home/sourabh/install/release/trunk-gcc/
Terminal:
```
>llvm-dev>>>sourabh@llvm-slrz2:~/work/dwarf/c_c++$ tot-gcc --version
gcc (GCC) 12.0.0 20210514 (experimental)
>llvm-dev>>>sourabh@llvm-slrz2:~/work/dwarf/c_c++$ gdb --version
GNU gdb (GDB) 11.0.50.20210512-git
>llvm-dev>>>sourabh@llvm:~/work/dwarf/c_c++$ cat macro.c
#define HELLO_GDB 1
int main() {}
>llvm-dev>>>sourabh@llvm:~/work/dwarf/c_c++$ tot-gcc -gdwarf-5 -g3 macro.c -o a.out.gcc
>llvm-dev>>>sourabh@llvm:~/work/dwarf/c_c++$ gdb a.out.gcc -q
Reading symbols from a.out.gcc...
(gdb) start
Temporary breakpoint 1 at 0x40048f: file macro.c, line 2.
Starting program: /home/sourabh/work/dwarf/c_c++/a.out
Temporary breakpoint 1, main () at macro.c:2
2 int main() {}
(gdb) info macros
>>> main_file->filename -> // printf
sal.symtab->filename -> macro.c // printf
(gdb)
>llvm-dev>>>sourabh@llvm-slrz2:~/work/dwarf/c_c++$ llvm-dwarfdump --debug-line a.out.gcc
a.out.gcc: file format elf64-x86-64
.debug_line contents:
debug_line[0x00000000]
Line table prologue:
total_length: 0x00000058
format: DWARF32
**version: 2 //** Version is 2, even at -gdwarf-5 -gstrict-dwarf, is this Okay** ?
prologue_length: 0x0000003c
min_inst_length: 1
default_is_stmt: 1
line_base: -5
line_range: 14
opcode_base: 13
include_directories[ 1] = "/usr/include"
file_names[ 1]:
name: "macro.c"
dir_index: 0
mod_time: 0x00000000
length: 0x00000000
file_names[ 2]:
name: "stdc-predef.h"
dir_index: 1
mod_time: 0x00000000
length: 0x00000000
```
Please let me know if above info suffices the details necessary ?
>On my side, I see this with gcc 11 (from the Ubuntu 21.04 package):
> root@bc7daa9a3ffc:/# gcc-11 --version
> gcc-11 (Ubuntu 11.1.0-1ubuntu1~21.04) 11.1.0
> root@bc7daa9a3ffc:/# cat test.c
> #include <stdio.h>
> #define FOO
> int main() {
> }
> root@bc7daa9a3ffc:/# gcc-11 -gdwarf-5 -g3 test.c
> root@bc7daa9a3ffc:/# llvm-dwarfdump-12 --debug-line a.out
...
> include_directories[ 0] = "/usr/include"
..
> file_names[ 0]:
> name: "test.c"
> dir_index: 0
> file_names[ 1]:
> name: "test.c"
> dir_index: 0
> file_names[ 2]:
> name: "stdc-predef.h"
> dir_index: 1
> file_names[ 3]:
> name: "stdio.h"
> dir_index: 1
>So from what I see, GCC 11 generates a dir[0] entry and a file[0] entry.
>However, the dir[0] value does not look good. It makes it appear as if my source file was at /usr/include/test.c, which it wasn't (it was at /test.c).
Thanks for sharing this, since both file0 and dir0 both are present, if possible, could you please share any behavior change or anything subtle you notice before/after
applying patch ?
Thanks,
Sourabh
-----Original Message-----
From: Simon Marchi <simon.marchi@polymtl.ca>
Sent: Saturday, May 15, 2021 12:20 AM
To: Tomar, Sourabh Singh <SourabhSingh.Tomar@amd.com>; Keith Seitz <keiths@redhat.com>; gdb-patches@sourceware.org
Subject: Re: [PATCH] Fix macro info lookup for binaries containing DWARFv5 line table
[CAUTION: External Email]
> One thing I noticed that changed is that GCC is now emitting `dir0`
> entry in line table as per DWARF5. However this still seems Incomplete to me:
> $llvm-dwarfdump --debug-line a.out
> ```
> include_directories[ 1] = "/usr/include" // index still starting from `1` even at "-gdwarf-5"
> file_names[ 1]:
> name: "macro.c"
> dir_index: 0 // -> dir 0 mentioned but dir 0 is not present in list of directories, i.e index still starting from 1
> mod_time: 0x00000000
> length: 0x00000000
> file_names[ 2]:
> name: "stdc-predef.h"
> dir_index: 1
> mod_time: 0x00000000
> length: 0x00000000
> ```
It's not clear how you produced this output (which compiler and which command). What's the version of that section? Because it looks like a DWARF 4 or before section. GCC 10 produces a DWARF 3 .debug_line section, even when using -gdwarf-5.
In DWARF 4 and before, a dir_index of 0 referred to the compilation directory, even if there wasn't an explicit dir[0] value. See section 6.2.4, bullet 12, of DWARF4.pdf. So you would look up the DW_AT_comp attribute of the compilation unit, append "macro.c", and that would give you the location of that file.
On my side, I see this with gcc 11 (from the Ubuntu 21.04 package):
root@bc7daa9a3ffc:/# gcc-11 --version
gcc-11 (Ubuntu 11.1.0-1ubuntu1~21.04) 11.1.0
root@bc7daa9a3ffc:/# cat test.c
#include <stdio.h>
#define FOO
int main() {
}
root@bc7daa9a3ffc:/# gcc-11 -gdwarf-5 -g3 test.c
root@bc7daa9a3ffc:/# llvm-dwarfdump-12 --debug-line a.out
...
include_directories[ 0] = "/usr/include"
include_directories[ 1] = "/usr/include"
include_directories[ 2] = "/usr/include/x86_64-linux-gnu/bits"
include_directories[ 3] = "/usr/include/x86_64-linux-gnu/sys"
include_directories[ 4] = "/usr/include/x86_64-linux-gnu/gnu"
include_directories[ 5] = "/usr/lib/gcc/x86_64-linux-gnu/11/include"
include_directories[ 6] = "/usr/include/x86_64-linux-gnu/bits/types"
file_names[ 0]:
name: "test.c"
dir_index: 0
file_names[ 1]:
name: "test.c"
dir_index: 0
file_names[ 2]:
name: "stdc-predef.h"
dir_index: 1
file_names[ 3]:
name: "stdio.h"
dir_index: 1
So from what I see, GCC 11 generates a dir[0] entry and a file[0] entry.
However, the dir[0] value does not look good. It makes it appear as if my source file was at /usr/include/test.c, which it wasn't (it was at /test.c).
Simon
next prev parent reply other threads:[~2021-05-14 19:20 UTC|newest]
Thread overview: 14+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-05-12 17:16 Sourabh Singh Tomar via Gdb-patches
2021-05-12 17:31 ` Tomar, Sourabh Singh via Gdb-patches
2021-05-13 15:47 ` Tom Tromey
2021-05-13 18:20 ` Simon Marchi via Gdb-patches
2021-05-14 14:56 ` Keith Seitz via Gdb-patches
2021-05-14 18:21 ` Tomar, Sourabh Singh via Gdb-patches
2021-05-14 18:50 ` Simon Marchi via Gdb-patches
2021-05-14 19:20 ` Tomar, Sourabh Singh via Gdb-patches [this message]
2021-05-14 20:56 ` Simon Marchi via Gdb-patches
2021-05-24 11:36 ` Tomar, Sourabh Singh via Gdb-patches
2021-05-24 18:47 ` Keith Seitz via Gdb-patches
2021-06-08 18:48 ` Keith Seitz via Gdb-patches
2021-06-22 17:01 ` [RFC] PING [Re: [PATCH] Fix macro info lookup for binaries containing DWARFv5 line table] Keith Seitz via Gdb-patches
2021-06-22 20:52 ` Simon Marchi via Gdb-patches
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=DM4PR12MB5295FE3DA86D7D26DD10F4D89D509@DM4PR12MB5295.namprd12.prod.outlook.com \
--to=gdb-patches@sourceware.org \
--cc=SourabhSingh.Tomar@amd.com \
--cc=keiths@redhat.com \
--cc=simon.marchi@polymtl.ca \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox