[RFA] Fix display of entry point address

[RFA] Fix display of entry point address

[Matthew Gretton-Dann]

[-- Attachment #1: Type: text/plain, Size: 1528 bytes --]

Hi,

Please can someone review the attached patch?

When displaying the entry point to a file (for instance as part of the
output of `info files') gdb prints the raw address it has calculated.
This is misleading on some platforms (notably on ARM when the entry
point is to a Thumb-state function) as the address will contain
additional information.

For example (from the break-entry.exp test):

(gdb) file /home/mgretton/gdb/build/gdb/testsuite/gdb.base/break-entry
Reading symbols from /home/mgretton/gdb/build/gdb/testsuite/gdb.base/break-entry...(no debugging symbols found)...done.
(gdb) info files
Symbols from "/home/mgretton/gdb/build/gdb/testsuite/gdb.base/break-entry".
Local exec file:
	`/home/mgretton/gdb/build/gdb/testsuite/gdb.base/break-entry', file type elf32-littlearm.
	Entry point: 0x8171

This is confusing as trying to set a breakpoint at 0x8171 causes a
warning like the following to be displayed:

(gdb) break *0x8171
warning: Breakpoint address adjusted from 0x00008171 to 0x00008170.
Breakpoint 1 at 0x8170

The attached patch changes gdb's behaviour to strip the arch specific
bits off the entry point before displaying it. 

A testsuite run shows no regressions, and the gdb.base/break-entry.exp
test now passes in its entirety.

Suggested ChangeLog entry:

2010-04-30  Matthew Gretton-Dann  <matthew.gretton-dann@arm.com>

	* exec.c (print_section_info): Display entry point without arch 
	specific parts.

Thanks,

Matt

-- 
Matthew Gretton-Dann
Principal Engineer - Tools, PD Software
ARM Limited

[-- Attachment #2: 1004-gdb-entry-point.patch --]
[-- Type: text/x-patch, Size: 1068 bytes --]

Index: gdb/exec.c
===================================================================
RCS file: /cvs/src/src/gdb/exec.c,v
retrieving revision 1.97
diff -u -p -u -p -r1.97 exec.c
--- gdb/exec.c	14 Apr 2010 17:26:11 -0000	1.97
+++ gdb/exec.c	30 Apr 2010 13:05:27 -0000
@@ -669,6 +669,7 @@ print_section_info (struct target_sectio
     {
       /* gcc-3.4 does not like the initialization in <p == t->sections_end>.  */
       bfd_vma displacement = 0;
+      bfd_vma entry_point;
 
       for (p = t->sections; p < t->sections_end; p++)
 	{
@@ -690,9 +691,11 @@ print_section_info (struct target_sectio
 	warning (_("Cannot find section for the entry point of %s.\n"),
 		 bfd_get_filename (abfd));
 
+      entry_point = gdbarch_addr_bits_remove (gdbarch, 
+					      bfd_get_start_address (abfd) 
+						+ displacement);
       printf_filtered (_("\tEntry point: %s\n"),
-		       paddress (gdbarch, (bfd_get_start_address (abfd)
-					   + displacement)));
+		       paddress (gdbarch, entry_point));
     }
   for (p = t->sections; p < t->sections_end; p++)
     {

Re: [RFA] Fix display of entry point address

[Doug Evans]

On Fri, Apr 30, 2010 at 9:01 AM, Matthew Gretton-Dann
<matthew.gretton-dann@arm.com> wrote:
> Hi,
>
> Please can someone review the attached patch?
>
> When displaying the entry point to a file (for instance as part of the
> output of `info files') gdb prints the raw address it has calculated.
> This is misleading on some platforms (notably on ARM when the entry
> point is to a Thumb-state function) as the address will contain
> additional information.
>
> For example (from the break-entry.exp test):
>
> (gdb) file /home/mgretton/gdb/build/gdb/testsuite/gdb.base/break-entry
> Reading symbols from /home/mgretton/gdb/build/gdb/testsuite/gdb.base/break-entry...(no debugging symbols found)...done.
> (gdb) info files
> Symbols from "/home/mgretton/gdb/build/gdb/testsuite/gdb.base/break-entry".
> Local exec file:
>        `/home/mgretton/gdb/build/gdb/testsuite/gdb.base/break-entry', file type elf32-littlearm.
>        Entry point: 0x8171
>
> This is confusing as trying to set a breakpoint at 0x8171 causes a
> warning like the following to be displayed:
>
> (gdb) break *0x8171
> warning: Breakpoint address adjusted from 0x00008171 to 0x00008170.
> Breakpoint 1 at 0x8170
>
> The attached patch changes gdb's behaviour to strip the arch specific
> bits off the entry point before displaying it.
>
> A testsuite run shows no regressions, and the gdb.base/break-entry.exp
> test now passes in its entirety.
>
> Suggested ChangeLog entry:
>
> 2010-04-30  Matthew Gretton-Dann  <matthew.gretton-dann@arm.com>
>
>        * exec.c (print_section_info): Display entry point without arch
>        specific parts.
>
> Thanks,
>
> Matt
>
> --
> Matthew Gretton-Dann
> Principal Engineer - Tools, PD Software
> ARM Limited
>

Hi.  I'd give it a day to see if someone else comments, but the patch
is ok to me.

Re: [RFA] Fix display of entry point address

[Joel Brobecker]

> Hi.  I'd give it a day to see if someone else comments, but the patch
> is ok to me.

FWIW: This change makes sense to me too.

-- 
Joel