[RFC] Add support for .debug_gdb_scripts to mach-o.

[RFC] Add support for .debug_gdb_scripts to mach-o.

[Doug Evans]

Hi.

I have no means of testing this.
And it's not clear that adding .debug_gdb_scripts to dwarf_section_names_xlat
is reasonable.

Can anyone using mach-o test this, or suggest a better fix?
There is a testcase for .debug_gdb_scripts in
gdb/testsuite/gdb.python/py-section-script.exp
but I'm not sure it'll work as is on mach-o.

TIA

2011-12-27  Doug Evans  <dje@google.com>

	* mach-o.c (dwarf_section_names_xlat): Add .debug_gdb_scripts.

Index: mach-o.c
===================================================================
RCS file: /cvs/src/src/bfd/mach-o.c,v
retrieving revision 1.81
diff -u -p -r1.81 mach-o.c
--- mach-o.c	19 Dec 2011 15:42:36 -0000	1.81
+++ mach-o.c	28 Dec 2011 00:15:14 -0000
@@ -195,6 +195,9 @@ static const mach_o_section_name_xlat dw
     {	".debug_macro",			"__debug_macro",
 	SEC_DEBUGGING,			BFD_MACH_O_S_REGULAR,
 	BFD_MACH_O_S_ATTR_DEBUG,	0},
+    {	".debug_gdb_scripts",		"__debug_gdb_scr",
+	SEC_DEBUGGING,			BFD_MACH_O_S_REGULAR,
+	BFD_MACH_O_S_ATTR_DEBUG,	0},
     { NULL, NULL, 0, 0, 0, 0}
   };
 

Re: [RFC] Add support for .debug_gdb_scripts to mach-o.

[Iain Sandoe]

Hi Doug,

On 28 Dec 2011, at 00:23, Doug Evans wrote:
> I have no means of testing this.

bfd/binutils is pretty good with i686-apple-darwin9 as a target (and  
gas is gradually getting there);
.. gdb also builds for that target with the mods noted below.

> And it's not clear that adding .debug_gdb_scripts to  
> dwarf_section_names_xlat
> is reasonable.

It seems the right place to me...

> Can anyone using mach-o test this,

.. in its own right, it builds, but needs the following to add it to  
GAS:

diff --git a/gas/config/obj-macho.c b/gas/config/obj-macho.c
index 74fb0c9..97762f8 100644
--- a/gas/config/obj-macho.c
+++ b/gas/config/obj-macho.c
@@ -526,7 +526,8 @@ static const char * const debug_sections[] =
    /*  9 */ ".debug_pubtypes",
    /* 10 */ ".debug_str",
    /* 11 */ ".debug_ranges",
-  /* 12 */ ".debug_macro"
+  /* 12 */ ".debug_macro",
+  /* 13 */ ".debug_gdb_scripts"
  };

  /* ??? Maybe these should be conditional on gdwarf-*.
@@ -823,6 +824,7 @@ const pseudo_typeS mach_o_pseudo_table[] =
    { "debug_str", obj_mach_o_debug_section, 10}, /* extension.  */
    { "debug_ranges", obj_mach_o_debug_section, 11}, /* extension.  */
    { "debug_macro", obj_mach_o_debug_section, 12}, /* extension.  */
+  { "debug_gdb_scripts", obj_mach_o_debug_section, 13}, /*  
extension.  */

    { "lazy_symbol_pointer", obj_mach_o_opt_tgt_section, 1},
    { "lazy_symbol_pointer2", obj_mach_o_opt_tgt_section, 2}, /*  
extension.  */


> There is a testcase for .debug_gdb_scripts in
> gdb/testsuite/gdb.python/py-section-script.exp
> but I'm not sure it'll work as is on mach-o.

no, it won't - we don't yet support .push/pop section in mach-o gas.

I suppose we should .. but not sure when it will hit the top of the  
TODO ;)

===

However, in the interim, Darwin's [native] ld supports creating a  
section from a file.
  (so there's a non-assembler way of doing this too).

from darwin's man ld:

    Options that control additional content
      -sectcreate segname sectname file
                  The section sectname in the segment segname is  
created from the contents of
                  file file. The combination of segname and sectname  
must be unique D there
                  cannot already be a section (segname,sectname) from  
any other input.

... so the following ought to work  ...

ld -sectcreate __DWARF __debug_gdb_scr python-script-file.py

(will try to fit in a look at this, at some stage - one would need http://sourceware.org/ml/gdb-patches/2011-12/msg00890.html 
  as well, to make it build).

cheers
Iain

Re: [RFC] Add support for .debug_gdb_scripts to mach-o.

[Iain Sandoe]

On 31 Dec 2011, at 15:12, Iain Sandoe wrote:

>> There is a testcase for .debug_gdb_scripts in
>> gdb/testsuite/gdb.python/py-section-script.exp
>> but I'm not sure it'll work as is on mach-o.
>
> no, it won't - we don't yet support .push/pop section in mach-o gas.

Alas, although that wasn't too hard to implement, we're still some way  
away from being able to build GCC with GAS, so we still need to work  
with the native system tools.

> (will try to fit in a look at this, at some stage

With http://sourceware.org/ml/gdb-patches/2011-12/msg00890.html  
applied, the following patch below gives:

Native configuration is i686-apple-darwin9

                 === gdb tests ===
...
Running ../../../src-git/gdb/testsuite/gdb.python/py-section- 
script.exp ...

                 === gdb Summary ===

# of expected passes            6

====

NOTES:

1. We cannot put a 'debug' section attribute on the debug_gdb_scripts  
section, because then dsymutil will automatically strip it from the  
exe - but (presumably because it's unrecognized) it doesn't put it  
into the dSYM, even when one specifies 'no_dead_strip'
.. :-( ...

2. The section switch has to be done specifically until we can build  
with gas.

3. maybe the DEFINE_GDB_SCRIPT macro should be guarded something like...
#if defined (__ELF__)
...
#elif defined (__MACH__)
...
(but I'm not sure how many targets can use it as stands...)

cheers
Iain

====


diff --git a/bfd/mach-o.c b/bfd/mach-o.c
index cc68d89..1404bb9 100644
--- a/bfd/mach-o.c
+++ b/bfd/mach-o.c
@@ -195,6 +195,9 @@ static const mach_o_section_name_xlat  
dwarf_section_names_xlat[] =
      {	".debug_macro",			"__debug_macro",
  	SEC_DEBUGGING,			BFD_MACH_O_S_REGULAR,
  	BFD_MACH_O_S_ATTR_DEBUG,	0},
+    {	".debug_gdb_scripts",		"__debug_gdb_scr",
+	SEC_DEBUGGING,			BFD_MACH_O_S_REGULAR,
+	BFD_MACH_O_S_ATTR_NO_DEAD_STRIP, 0},
      { NULL, NULL, 0, 0, 0, 0}
    };


diff --git a/gas/config/obj-macho.c b/gas/config/obj-macho.c
index 804d42b..4953a91 100644
--- a/gas/config/obj-macho.c
+++ b/gas/config/obj-macho.c
@@ -559,7 +561,8 @@ static const char * const debug_sections[] =
    /*  9 */ ".debug_pubtypes",
    /* 10 */ ".debug_str",
    /* 11 */ ".debug_ranges",
-  /* 12 */ ".debug_macro"
+  /* 12 */ ".debug_macro",
+  /* 13 */ ".debug_gdb_scripts",
  };

  /* ??? Maybe these should be conditional on gdwarf-*.
@@ -1064,6 +1067,7 @@ const pseudo_typeS mach_o_pseudo_table[] =
    { "debug_str", obj_mach_o_debug_section, 10}, /* extension.  */
    { "debug_ranges", obj_mach_o_debug_section, 11}, /* extension.  */
    { "debug_macro", obj_mach_o_debug_section, 12}, /* extension.  */
+  { "debug_gdb_scripts", obj_mach_o_debug_section, 13}, /*  
extension.  */

    { "lazy_symbol_pointer", obj_mach_o_opt_tgt_section, 1},
    { "lazy_symbol_pointer2", obj_mach_o_opt_tgt_section, 2}, /*  
extension.  */

diff --git a/gdb/testsuite/gdb.python/py-section-script.c b/gdb/ 
testsuite/gdb.python/py-section-script.c
index 1ead7dd..6d48071 100644
--- a/gdb/testsuite/gdb.python/py-section-script.c
+++ b/gdb/testsuite/gdb.python/py-section-script.c
@@ -18,6 +18,7 @@
  /* Put the path to the pretty-printer script in .debug_gdb_scripts so
     gdb will automagically loaded it.  */

+#ifndef __MACH__
  #define DEFINE_GDB_SCRIPT(script_name) \
    asm("\
  .pushsection \".debug_gdb_scripts\", \"MS\",@progbits,1\n\
@@ -25,6 +26,14 @@
  .asciz \"" script_name "\"\n\
  .popsection \n\
  ");
+#else
+#define DEFINE_GDB_SCRIPT(script_name) \
+  asm(".section __DWARF,__debug_gdb_scr,regular,no_dead_strip\n\
+	.byte 1\n\
+	.asciz \"" script_name "\"\n\
+	.text\n\
+      ");
+#endif

  DEFINE_GDB_SCRIPT ("py-section-script.py")


diff --git a/gdb/testsuite/gdb.python/py-section-script.exp b/gdb/ 
testsuite/gdb.python/py-section-script.exp
index 31fdcf8..1cde576 100644
--- a/gdb/testsuite/gdb.python/py-section-script.exp
+++ b/gdb/testsuite/gdb.python/py-section-script.exp
@@ -24,7 +24,8 @@ if {![istarget *-*-linux*]
      && ![istarget *-*-openbsd*]
      && ![istarget arm*-*-eabi*]
      && ![istarget arm*-*-symbianelf*]
-    && ![istarget powerpc-*-eabi*]} {
+    && ![istarget powerpc-*-eabi*]
+    && ![istarget i?86-*-darwin*]} {
      verbose "Skipping py-section-script.exp because of lack of  
support."
      return
  }