[v2,4/5] Allow display of negative offsets in print_address_symbolic()

Message ID 20190704045503.1250-5-kevinb@redhat.com
State New
Headers show
Series
  • Non-contiguous address range bug fixes / improvements
Related show

Commit Message

Kevin Buettner July 4, 2019, 4:55 a.m.
When examining addresses associated with blocks with non-contiguous
address ranges, it's not uncommon to see large positive offsets which,
for some address width, actually represent a smaller negative offset.
Here's an example taken from the test case (using the
dw2-ranges-func-lo-cold executable):

    (gdb) x/5i foo_cold
       0x40110d <foo+4294967277>:	push   %rbp
       0x40110e <foo+4294967278>:	mov    %rsp,%rbp
       0x401111 <foo+4294967281>:	callq  0x401106 <baz>
       0x401116 <foo+4294967286>:	nop
       0x401117 <foo+4294967287>:	pop    %rbp

This commit, in conjuction with an earlier patch from this series, causes
cases like the above to be displayed like this (below) instead:

(gdb) x/5i foo_cold
   0x40110d <foo_cold>:	push   %rbp
   0x40110e <foo-18>:	mov    %rsp,%rbp
   0x401111 <foo-15>:	callq  0x401106 <baz>
   0x401116 <foo-10>:	nop
   0x401117 <foo-9>:	pop    %rbp

Note that the address of foo_cold is now (due to another patch) being
displayed as <foo_cold> instead of <foo+BigOffset>.  The subsequent
lines are shown as negative offsets from foo.

Disassembly using the "disassemble" command is somewhat affected by
these changes:

Before:

(gdb) disassemble foo_cold
Dump of assembler code for function foo:
Address range 0x401120 to 0x40113b:
   0x0000000000401120 <+0>:	push   %rbp
   0x0000000000401121 <+1>:	mov    %rsp,%rbp
   0x0000000000401124 <+4>:	callq  0x401119 <bar>
   0x0000000000401129 <+9>:	mov    0x2ef1(%rip),%eax        # 0x404020 <e>
   0x000000000040112f <+15>:	test   %eax,%eax
   0x0000000000401131 <+17>:	je     0x401138 <foo+24>
   0x0000000000401133 <+19>:	callq  0x40110d <foo+4294967277>
   0x0000000000401138 <+24>:	nop
   0x0000000000401139 <+25>:	pop    %rbp
   0x000000000040113a <+26>:	retq
Address range 0x40110d to 0x401119:
   0x000000000040110d <+-19>:	push   %rbp
   0x000000000040110e <+-18>:	mov    %rsp,%rbp
   0x0000000000401111 <+-15>:	callq  0x401106 <baz>
   0x0000000000401116 <+-10>:	nop
   0x0000000000401117 <+-9>:	pop    %rbp
   0x0000000000401118 <+-8>:	retq
End of assembler dump.

After:

(gdb) disassemble foo_cold
Dump of assembler code for function foo:
Address range 0x401120 to 0x40113b:
   0x0000000000401120 <+0>:	push   %rbp
   0x0000000000401121 <+1>:	mov    %rsp,%rbp
   0x0000000000401124 <+4>:	callq  0x401119 <bar>
   0x0000000000401129 <+9>:	mov    0x2ef1(%rip),%eax        # 0x404020 <e>
   0x000000000040112f <+15>:	test   %eax,%eax
   0x0000000000401131 <+17>:	je     0x401138 <foo+24>
   0x0000000000401133 <+19>:	callq  0x40110d <foo_cold>
   0x0000000000401138 <+24>:	nop
   0x0000000000401139 <+25>:	pop    %rbp
   0x000000000040113a <+26>:	retq
Address range 0x40110d to 0x401119:
   0x000000000040110d <-19>:	push   %rbp
   0x000000000040110e <-18>:	mov    %rsp,%rbp
   0x0000000000401111 <-15>:	callq  0x401106 <baz>
   0x0000000000401116 <-10>:	nop
   0x0000000000401117 <-9>:	pop    %rbp
   0x0000000000401118 <-8>:	retq
End of assembler dump.

Note that negative offsets are now displayed without the leading "+".
Also, the callq to foo_cold is now displayed as such instead of a callq
to foo with a large positive offset.

gdb/ChangeLog:

	* printcmd.c (print_address_symbolic): Print negative offsets.
	(build_address_symbolic): Force signed arithmetic when computing
	offset.
---
 gdb/printcmd.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

-- 
2.21.0

Comments

Kevin Buettner July 30, 2019, 4:47 p.m. | #1
I somehow missed pushing this commit when I pushed the rest of the
series.  It was strange though - that second hunk of the patch was in,
but the first was not.  Moreover, I didn't see an actual commit in master.

Regardless, it should be in now.

Thanks to Tom de Vries for his help in finding this problem.

Kevin

On Wed,  3 Jul 2019 21:55:02 -0700
Kevin Buettner <kevinb@redhat.com> wrote:

> When examining addresses associated with blocks with non-contiguous

> address ranges, it's not uncommon to see large positive offsets which,

> for some address width, actually represent a smaller negative offset.

> Here's an example taken from the test case (using the

> dw2-ranges-func-lo-cold executable):

> 

>     (gdb) x/5i foo_cold

>        0x40110d <foo+4294967277>:	push   %rbp

>        0x40110e <foo+4294967278>:	mov    %rsp,%rbp

>        0x401111 <foo+4294967281>:	callq  0x401106 <baz>

>        0x401116 <foo+4294967286>:	nop

>        0x401117 <foo+4294967287>:	pop    %rbp

> 

> This commit, in conjuction with an earlier patch from this series, causes

> cases like the above to be displayed like this (below) instead:

> 

> (gdb) x/5i foo_cold

>    0x40110d <foo_cold>:	push   %rbp

>    0x40110e <foo-18>:	mov    %rsp,%rbp

>    0x401111 <foo-15>:	callq  0x401106 <baz>

>    0x401116 <foo-10>:	nop

>    0x401117 <foo-9>:	pop    %rbp

> 

> Note that the address of foo_cold is now (due to another patch) being

> displayed as <foo_cold> instead of <foo+BigOffset>.  The subsequent

> lines are shown as negative offsets from foo.

> 

> Disassembly using the "disassemble" command is somewhat affected by

> these changes:

> 

> Before:

> 

> (gdb) disassemble foo_cold

> Dump of assembler code for function foo:

> Address range 0x401120 to 0x40113b:

>    0x0000000000401120 <+0>:	push   %rbp

>    0x0000000000401121 <+1>:	mov    %rsp,%rbp

>    0x0000000000401124 <+4>:	callq  0x401119 <bar>

>    0x0000000000401129 <+9>:	mov    0x2ef1(%rip),%eax        # 0x404020 <e>

>    0x000000000040112f <+15>:	test   %eax,%eax

>    0x0000000000401131 <+17>:	je     0x401138 <foo+24>

>    0x0000000000401133 <+19>:	callq  0x40110d <foo+4294967277>

>    0x0000000000401138 <+24>:	nop

>    0x0000000000401139 <+25>:	pop    %rbp

>    0x000000000040113a <+26>:	retq

> Address range 0x40110d to 0x401119:

>    0x000000000040110d <+-19>:	push   %rbp

>    0x000000000040110e <+-18>:	mov    %rsp,%rbp

>    0x0000000000401111 <+-15>:	callq  0x401106 <baz>

>    0x0000000000401116 <+-10>:	nop

>    0x0000000000401117 <+-9>:	pop    %rbp

>    0x0000000000401118 <+-8>:	retq

> End of assembler dump.

> 

> After:

> 

> (gdb) disassemble foo_cold

> Dump of assembler code for function foo:

> Address range 0x401120 to 0x40113b:

>    0x0000000000401120 <+0>:	push   %rbp

>    0x0000000000401121 <+1>:	mov    %rsp,%rbp

>    0x0000000000401124 <+4>:	callq  0x401119 <bar>

>    0x0000000000401129 <+9>:	mov    0x2ef1(%rip),%eax        # 0x404020 <e>

>    0x000000000040112f <+15>:	test   %eax,%eax

>    0x0000000000401131 <+17>:	je     0x401138 <foo+24>

>    0x0000000000401133 <+19>:	callq  0x40110d <foo_cold>

>    0x0000000000401138 <+24>:	nop

>    0x0000000000401139 <+25>:	pop    %rbp

>    0x000000000040113a <+26>:	retq

> Address range 0x40110d to 0x401119:

>    0x000000000040110d <-19>:	push   %rbp

>    0x000000000040110e <-18>:	mov    %rsp,%rbp

>    0x0000000000401111 <-15>:	callq  0x401106 <baz>

>    0x0000000000401116 <-10>:	nop

>    0x0000000000401117 <-9>:	pop    %rbp

>    0x0000000000401118 <-8>:	retq

> End of assembler dump.

> 

> Note that negative offsets are now displayed without the leading "+".

> Also, the callq to foo_cold is now displayed as such instead of a callq

> to foo with a large positive offset.

> 

> gdb/ChangeLog:

> 

> 	* printcmd.c (print_address_symbolic): Print negative offsets.

> 	(build_address_symbolic): Force signed arithmetic when computing

> 	offset.

> ---

>  gdb/printcmd.c | 4 ++--

>  1 file changed, 2 insertions(+), 2 deletions(-)

> 

> diff --git a/gdb/printcmd.c b/gdb/printcmd.c

> index 1109cb3046..dce6ab2db9 100644

> --- a/gdb/printcmd.c

> +++ b/gdb/printcmd.c

> @@ -539,7 +539,7 @@ print_address_symbolic (struct gdbarch *gdbarch, CORE_ADDR addr,

>      fputs_filtered ("<", stream);

>    fputs_styled (name.c_str (), function_name_style.style (), stream);

>    if (offset != 0)

> -    fprintf_filtered (stream, "+%u", (unsigned int) offset);

> +    fprintf_filtered (stream, "%+d", offset);

>  

>    /* Append source filename and line number if desired.  Give specific

>       line # of this addr, if we have it; else line # of the nearest symbol.  */

> @@ -679,7 +679,7 @@ build_address_symbolic (struct gdbarch *gdbarch,

>        && name_location + max_symbolic_offset > name_location)

>      return 1;

>  

> -  *offset = addr - name_location;

> +  *offset = (LONGEST) addr - name_location;

>  

>    *name = name_temp;

>  

> -- 

> 2.21.0

>

Patch

diff --git a/gdb/printcmd.c b/gdb/printcmd.c
index 1109cb3046..dce6ab2db9 100644
--- a/gdb/printcmd.c
+++ b/gdb/printcmd.c
@@ -539,7 +539,7 @@  print_address_symbolic (struct gdbarch *gdbarch, CORE_ADDR addr,
     fputs_filtered ("<", stream);
   fputs_styled (name.c_str (), function_name_style.style (), stream);
   if (offset != 0)
-    fprintf_filtered (stream, "+%u", (unsigned int) offset);
+    fprintf_filtered (stream, "%+d", offset);
 
   /* Append source filename and line number if desired.  Give specific
      line # of this addr, if we have it; else line # of the nearest symbol.  */
@@ -679,7 +679,7 @@  build_address_symbolic (struct gdbarch *gdbarch,
       && name_location + max_symbolic_offset > name_location)
     return 1;
 
-  *offset = addr - name_location;
+  *offset = (LONGEST) addr - name_location;
 
   *name = name_temp;