[RFC/RFA] PING: skip __main

[RFC/RFA] PING: skip __main

[Pierre Muller]

  Pedro submitted this patch end of January,
but the 6.8 release pushed it to a later date.
http://sourceware.org/ml/gdb-patches/2008-01/msg00665.html
  
Now that GDB 6.8 is out, 
I would really like to get this patch in.
Pedro, this is your work, but you seem to be
much more involved in other things lately,
thus, if you don't mind, I can try to push this through. 

  I updated the patch from the first message
referenced and just followed Daniel's advice in
http://sourceware.org/ml/gdb-patches/2008-01/msg00883.html
to change skip___main into skip_main_constructor.

  This __main call seems to be
a common feature for coff format, thus 
I suppose that it applies to other targets,
go32v2 djgpp is probably one of them.

For cygwin target, I get this:

                === gdb Summary ===
	
-# of expected passes           10733
-# of unexpected failures       560
+# of expected passes           10928
+# of unexpected failures       363
 # of expected failures         59
 # of unknown successes         2
-# of known failures            21
+# of known failures            23
 # of unresolved testcases      40
 # of untested testcases                14
 # of unsupported tests         23

  Almost 200 failures less...

  Daniel wanted to get a comment from Mark
in his last email in that thread, that is the reason
why I added Mark Kettenis to the list of recipients.

Does this patch look OK now?

  One point that should still be discussed is
if we should call main_name function
rather than hardcoded "main" in find_function_start_sal in symtab.c.

Pierre Muller
Pascal language support maintainer for GDB


2008-05-02  Pedro Alves  <pedro_alves@portugalmail.pt>
	Pierre Muller  <muller@ics.u-strasbg.fr>

	* gdbarch.sh (gdbarch_skip_main_constructor_call): New.
	* gdbarch.h, gdbarch.c: Regenerate.

	* i386-tdep.h (i386_skip_main_constructor_call): Declare.
	* i386-tdep.c (i386_skip_main_constructor_call): New.
	* i386-cygwin-tdep.c (i386_cygwin_init_abi): Register
	i386_skip_main_constructor_call as
gdbarch_skip_main_constructor_call 
	gdbarch callback.
	* symtab.c (find_function_start_sal): When pc points at the "main"
	function, call gdbarch_skip_main_constructor_call.

Index: gdb/gdbarch.c
===================================================================
RCS file: /cvs/src/src/gdb/gdbarch.c,v
retrieving revision 1.422
diff -u -p -r1.422 gdbarch.c
--- gdb/gdbarch.c	22 Apr 2008 11:03:41 -0000	1.422
+++ gdb/gdbarch.c	30 Apr 2008 13:01:33 -0000
@@ -184,6 +184,7 @@ struct gdbarch
   gdbarch_integer_to_address_ftype *integer_to_address;
   gdbarch_return_value_ftype *return_value;
   gdbarch_skip_prologue_ftype *skip_prologue;
+  gdbarch_skip_main_constructor_call_ftype *skip_main_constructor_call;
   gdbarch_inner_than_ftype *inner_than;
   gdbarch_breakpoint_from_pc_ftype *breakpoint_from_pc;
   gdbarch_adjust_breakpoint_address_ftype *adjust_breakpoint_address;
@@ -306,6 +307,7 @@ struct gdbarch startup_gdbarch =
   0,  /* integer_to_address */
   0,  /* return_value */
   0,  /* skip_prologue */
+  0,  /* skip_main_constructor_call */
   0,  /* inner_than */
   0,  /* breakpoint_from_pc */
   0,  /* adjust_breakpoint_address */
@@ -542,6 +544,7 @@ verify_gdbarch (struct gdbarch *gdbarch)
   /* Skip verify of return_value, has predicate */
   if (gdbarch->skip_prologue == 0)
     fprintf_unfiltered (log, "\n\tskip_prologue");
+  /* Skip verify of skip_main_constructor_call, has predicate */
   if (gdbarch->inner_than == 0)
     fprintf_unfiltered (log, "\n\tinner_than");
   if (gdbarch->breakpoint_from_pc == 0)
@@ -934,6 +937,12 @@ gdbarch_dump (struct gdbarch *gdbarch, s
                       "gdbarch_dump: single_step_through_delay =
<0x%lx>\n",
                       (long) gdbarch->single_step_through_delay);
   fprintf_unfiltered (file,
+                      "gdbarch_dump: gdbarch_skip_main_constructor_call_p()
= %d\n",
+                      gdbarch_skip_main_constructor_call_p (gdbarch));
+  fprintf_unfiltered (file,
+                      "gdbarch_dump: skip_main_constructor_call =
<0x%lx>\n",
+                      (long) gdbarch->skip_main_constructor_call);
+  fprintf_unfiltered (file,
                       "gdbarch_dump: gdbarch_skip_permanent_breakpoint_p()
= %d\n",
                       gdbarch_skip_permanent_breakpoint_p (gdbarch));
   fprintf_unfiltered (file,
@@ -2075,6 +2084,30 @@ set_gdbarch_skip_prologue (struct gdbarc
 }
 
 int
+gdbarch_skip_main_constructor_call_p (struct gdbarch *gdbarch)
+{
+  gdb_assert (gdbarch != NULL);
+  return gdbarch->skip_main_constructor_call != NULL;
+}
+
+CORE_ADDR
+gdbarch_skip_main_constructor_call (struct gdbarch *gdbarch, CORE_ADDR ip)
+{
+  gdb_assert (gdbarch != NULL);
+  gdb_assert (gdbarch->skip_main_constructor_call != NULL);
+  if (gdbarch_debug >= 2)
+    fprintf_unfiltered (gdb_stdlog, "gdbarch_skip_main_constructor_call
called\n");
+  return gdbarch->skip_main_constructor_call (gdbarch, ip);
+}
+
+void
+set_gdbarch_skip_main_constructor_call (struct gdbarch *gdbarch,
+                              gdbarch_skip_main_constructor_call_ftype
skip_main_constructor_call)
+{
+  gdbarch->skip_main_constructor_call = skip_main_constructor_call;
+}
+
+int
 gdbarch_inner_than (struct gdbarch *gdbarch, CORE_ADDR lhs, CORE_ADDR rhs)
 {
   gdb_assert (gdbarch != NULL);
Index: gdb/gdbarch.h
===================================================================
RCS file: /cvs/src/src/gdb/gdbarch.h,v
retrieving revision 1.377
diff -u -p -r1.377 gdbarch.h
--- gdb/gdbarch.h	29 Apr 2008 16:06:07 -0000	1.377
+++ gdb/gdbarch.h	30 Apr 2008 13:01:33 -0000
@@ -381,6 +381,12 @@ typedef CORE_ADDR (gdbarch_skip_prologue
 extern CORE_ADDR gdbarch_skip_prologue (struct gdbarch *gdbarch, CORE_ADDR
ip);
 extern void set_gdbarch_skip_prologue (struct gdbarch *gdbarch,
gdbarch_skip_prologue_ftype *skip_prologue);
 
+extern int gdbarch_skip_main_constructor_call_p (struct gdbarch *gdbarch);
+
+typedef CORE_ADDR (gdbarch_skip_main_constructor_call_ftype) (struct
gdbarch *gdbarch, CORE_ADDR ip);
+extern CORE_ADDR gdbarch_skip_main_constructor_call (struct gdbarch
*gdbarch, CORE_ADDR ip);
+extern void set_gdbarch_skip_main_constructor_call (struct gdbarch
*gdbarch, gdbarch_skip_main_constructor_call_ftype
*skip_main_constructor_call);
+
 typedef int (gdbarch_inner_than_ftype) (CORE_ADDR lhs, CORE_ADDR rhs);
 extern int gdbarch_inner_than (struct gdbarch *gdbarch, CORE_ADDR lhs,
CORE_ADDR rhs);
 extern void set_gdbarch_inner_than (struct gdbarch *gdbarch,
gdbarch_inner_than_ftype *inner_than);
Index: gdb/gdbarch.sh
===================================================================
RCS file: /cvs/src/src/gdb/gdbarch.sh,v
retrieving revision 1.463
diff -u -p -r1.463 gdbarch.sh
--- gdb/gdbarch.sh	29 Apr 2008 16:06:06 -0000	1.463
+++ gdb/gdbarch.sh	30 Apr 2008 13:01:34 -0000
@@ -483,6 +483,7 @@ M:CORE_ADDR:integer_to_address:struct ty
 M:enum return_value_convention:return_value:struct type *functype, struct
type *valtype, struct regcache *regcache, gdb_byte *readbuf, const gdb_byte
*writebuf:functype, valtype, regcache, readbuf, writebuf
 
 m:CORE_ADDR:skip_prologue:CORE_ADDR ip:ip:0:0
+M:CORE_ADDR:skip_main_constructor_call:CORE_ADDR ip:ip
 f:int:inner_than:CORE_ADDR lhs, CORE_ADDR rhs:lhs, rhs:0:0
 m:const gdb_byte *:breakpoint_from_pc:CORE_ADDR *pcptr, int *lenptr:pcptr,
lenptr::0:
 M:CORE_ADDR:adjust_breakpoint_address:CORE_ADDR bpaddr:bpaddr
Index: gdb/i386-cygwin-tdep.c
===================================================================
RCS file: /cvs/src/src/gdb/i386-cygwin-tdep.c,v
retrieving revision 1.16
diff -u -p -r1.16 i386-cygwin-tdep.c
--- gdb/i386-cygwin-tdep.c	1 Jan 2008 22:53:10 -0000	1.16
+++ gdb/i386-cygwin-tdep.c	30 Apr 2008 13:01:34 -0000
@@ -227,6 +227,8 @@ i386_cygwin_init_abi (struct gdbarch_inf
 
   set_gdbarch_skip_trampoline_code (gdbarch,
i386_cygwin_skip_trampoline_code);
 
+  set_gdbarch_skip_main_constructor_call (gdbarch,
i386_skip_main_constructor_call);
+
   tdep->struct_return = reg_struct_return;
 
   tdep->gregset_reg_offset = i386_win32_gregset_reg_offset;
Index: gdb/i386-tdep.c
===================================================================
RCS file: /cvs/src/src/gdb/i386-tdep.c,v
retrieving revision 1.255
diff -u -p -r1.255 i386-tdep.c
--- gdb/i386-tdep.c	25 Apr 2008 14:57:30 -0000	1.255
+++ gdb/i386-tdep.c	30 Apr 2008 13:01:35 -0000
@@ -941,6 +941,33 @@ i386_skip_prologue (struct gdbarch *gdba
   return pc;
 }
 
+/* Check that the code pointed to by PC corresponds to a call to
+   __main, skip it if so.  Return PC otherwise.  */
+
+CORE_ADDR
+i386_skip_main_constructor_call (struct gdbarch *gdbarch, CORE_ADDR pc)
+{
+  gdb_byte op;
+
+  target_read_memory (pc, &op, 1);
+  if (op == 0xe8)
+    {
+      gdb_byte buf[4];
+      if (target_read_memory (pc + 1, buf, sizeof buf) == 0)
+ 	{
+ 	  CORE_ADDR call_dest = pc + 5 + extract_unsigned_integer (buf, 4);
+
+ 	  struct minimal_symbol *s = lookup_minimal_symbol_by_pc
(call_dest);
+ 	  if (s != NULL
+ 	      && SYMBOL_LINKAGE_NAME (s) != NULL
+ 	      && strcmp (SYMBOL_LINKAGE_NAME (s), "__main") == 0)
+ 	    pc += 5;
+ 	}
+    }
+
+  return pc;
+}
+
 /* This function is 64-bit safe.  */
 
 static CORE_ADDR
Index: gdb/i386-tdep.h
===================================================================
RCS file: /cvs/src/src/gdb/i386-tdep.h,v
retrieving revision 1.53
diff -u -p -r1.53 i386-tdep.h
--- gdb/i386-tdep.h	11 Mar 2008 05:21:38 -0000	1.53
+++ gdb/i386-tdep.h	30 Apr 2008 13:01:35 -0000
@@ -166,6 +166,7 @@ extern struct type *i386_sse_type (struc
 
 /* Functions exported from i386-tdep.c.  */
 extern CORE_ADDR i386_pe_skip_trampoline_code (CORE_ADDR pc, char *name);
+extern CORE_ADDR i386_skip_main_constructor_call (struct gdbarch *gdbarch,
CORE_ADDR pc);
 
 /* Return the name of register REGNUM.  */
 extern char const *i386_register_name (struct gdbarch * gdbarch, int
regnum);
Index: gdb/symtab.c
===================================================================
RCS file: /cvs/src/src/gdb/symtab.c,v
retrieving revision 1.177
diff -u -p -r1.177 symtab.c
--- gdb/symtab.c	19 Apr 2008 11:39:50 -0000	1.177
+++ gdb/symtab.c	30 Apr 2008 13:01:36 -0000
@@ -2572,6 +2572,21 @@ find_function_start_sal (struct symbol *
       /* Recalculate the line number (might not be N+1).  */
       sal = find_pc_sect_line (pc, SYMBOL_BFD_SECTION (sym), 0);
     }
+
+  /* On targets with executable formats that don't have a concept of
+     constructors (ELF with .init has, PE doesn't), gcc emits a call
+     to `__main' in `main' between the prologue and before user
+     code.  */
+  if (funfirstline
+      && gdbarch_skip_main_constructor_call_p (current_gdbarch)
+      && SYMBOL_LINKAGE_NAME (sym)
+      && strcmp (SYMBOL_LINKAGE_NAME (sym), "main") == 0)
+    {
+      pc = gdbarch_skip_main_constructor_call (current_gdbarch, pc);
+      /* Recalculate the line number (might not be N+1).  */
+      sal = find_pc_sect_line (pc, SYMBOL_BFD_SECTION (sym), 0);
+    }
+
   sal.pc = pc;
 
   return sal;

Re: [RFC/RFA] PING: skip __main

[Pedro Alves]

A Friday 02 May 2008 14:25:44, Pierre Muller wrote:
>   Pedro submitted this patch end of January,
> but the 6.8 release pushed it to a later date.
> http://sourceware.org/ml/gdb-patches/2008-01/msg00665.html
>
> Now that GDB 6.8 is out,
> I would really like to get this patch in.
> Pedro, this is your work, but you seem to be
> much more involved in other things lately,
> thus, if you don't mind, I can try to push this through.
>

I don't mind at all.  I even thank you.  :-)

>   I updated the patch from the first message
> referenced and just followed Daniel's advice in
> http://sourceware.org/ml/gdb-patches/2008-01/msg00883.html
> to change skip___main into skip_main_constructor.
>
>   This __main call seems to be
> a common feature for coff format, thus
> I suppose that it applies to other targets,
> go32v2 djgpp is probably one of them.
>
> For cygwin target, I get this:
>
>                 === gdb Summary ===
>
> -# of expected passes           10733
> -# of unexpected failures       560
> +# of expected passes           10928
> +# of unexpected failures       363
>  # of expected failures         59
>  # of unknown successes         2
> -# of known failures            21
> +# of known failures            23
>  # of unresolved testcases      40
>  # of untested testcases                14
>  # of unsupported tests         23
>
>   Almost 200 failures less...
>
>   Daniel wanted to get a comment from Mark
> in his last email in that thread, that is the reason
> why I added Mark Kettenis to the list of recipients.
>
> Does this patch look OK now?
>
>   One point that should still be discussed is
> if we should call main_name function
> rather than hardcoded "main" in find_function_start_sal in symtab.c.
>

I believe "__main" is only emitted by gcc in the "main"
entry point.  I'd expect other gcc languages' runtimes
to have something like this somewhere in the startup routines:

 int
 main ()
 {
        initialize_my_lang_runtime ();
	mylang_entry_point ();
 }

In gcc the call is emitted in:

void
expand_main_function (void)
{
#if (defined(INVOKE__main)                              \
     || (!defined(HAS_INIT_SECTION)                     \
         && !defined(INIT_SECTION_ASM_OP)               \
         && !defined(INIT_ARRAY_SECTION_ASM_OP)))
  emit_library_call (init_one_libfunc (NAME__MAIN), LCT_NORMAL, VOIDmode, 0);
#endif
}

And that is called like so:

  /* If this function is `main', emit a call to `__main'
     to run global initializers, etc.  */
  if (DECL_NAME (current_function_decl)
      && MAIN_NAME_P (DECL_NAME (current_function_decl))
      && DECL_FILE_SCOPE_P (current_function_decl))
    expand_main_function ();

#define MAIN_NAME_P(NODE) \
    (IDENTIFIER_NODE_CHECK (NODE) == main_identifier_node)

#define main_identifier_node             global_trees[TI_MAIN_IDENTIFIER]

I can only find C variants and ADA setting main_identifier_node,
and it's always on "main".  So, the ADA case alone invalidates usage
of main_name in this case.

ada/utils.c:763:  main_identifier_node = get_identifier ("main");
c-common.c:3912:  main_identifier_node = get_identifier ("main");

gdb/ada-lang.c:
/* The name of the symbol to use to get the name of the main subprogram.  */
static const char ADA_MAIN_PROGRAM_SYMBOL_NAME[]
  = "__gnat_ada_main_program_name";

> Pierre Muller
> Pascal language support maintainer for GDB
>
>
> 2008-05-02  Pedro Alves  <pedro_alves@portugalmail.pt>
> 	Pierre Muller  <muller@ics.u-strasbg.fr>
>
> 	* gdbarch.sh (gdbarch_skip_main_constructor_call): New.
> 	* gdbarch.h, gdbarch.c: Regenerate.
>
> 	* i386-tdep.h (i386_skip_main_constructor_call): Declare.
> 	* i386-tdep.c (i386_skip_main_constructor_call): New.
> 	* i386-cygwin-tdep.c (i386_cygwin_init_abi): Register
> 	i386_skip_main_constructor_call as
> gdbarch_skip_main_constructor_call
> 	gdbarch callback.
> 	* symtab.c (find_function_start_sal): When pc points at the "main"
> 	function, call gdbarch_skip_main_constructor_call.
>
> Index: gdb/gdbarch.c
> ===================================================================
> RCS file: /cvs/src/src/gdb/gdbarch.c,v
> retrieving revision 1.422
> diff -u -p -r1.422 gdbarch.c
> --- gdb/gdbarch.c	22 Apr 2008 11:03:41 -0000	1.422
> +++ gdb/gdbarch.c	30 Apr 2008 13:01:33 -0000
> @@ -184,6 +184,7 @@ struct gdbarch
>    gdbarch_integer_to_address_ftype *integer_to_address;
>    gdbarch_return_value_ftype *return_value;
>    gdbarch_skip_prologue_ftype *skip_prologue;
> +  gdbarch_skip_main_constructor_call_ftype *skip_main_constructor_call;
>    gdbarch_inner_than_ftype *inner_than;
>    gdbarch_breakpoint_from_pc_ftype *breakpoint_from_pc;
>    gdbarch_adjust_breakpoint_address_ftype *adjust_breakpoint_address;
> @@ -306,6 +307,7 @@ struct gdbarch startup_gdbarch =
>    0,  /* integer_to_address */
>    0,  /* return_value */
>    0,  /* skip_prologue */
> +  0,  /* skip_main_constructor_call */
>    0,  /* inner_than */
>    0,  /* breakpoint_from_pc */
>    0,  /* adjust_breakpoint_address */
> @@ -542,6 +544,7 @@ verify_gdbarch (struct gdbarch *gdbarch)
>    /* Skip verify of return_value, has predicate */
>    if (gdbarch->skip_prologue == 0)
>      fprintf_unfiltered (log, "\n\tskip_prologue");
> +  /* Skip verify of skip_main_constructor_call, has predicate */
>    if (gdbarch->inner_than == 0)
>      fprintf_unfiltered (log, "\n\tinner_than");
>    if (gdbarch->breakpoint_from_pc == 0)
> @@ -934,6 +937,12 @@ gdbarch_dump (struct gdbarch *gdbarch, s
>                        "gdbarch_dump: single_step_through_delay =
> <0x%lx>\n",
>                        (long) gdbarch->single_step_through_delay);
>    fprintf_unfiltered (file,
> +                      "gdbarch_dump:
> gdbarch_skip_main_constructor_call_p() = %d\n",
> +                      gdbarch_skip_main_constructor_call_p (gdbarch));
> +  fprintf_unfiltered (file,
> +                      "gdbarch_dump: skip_main_constructor_call =
> <0x%lx>\n",
> +                      (long) gdbarch->skip_main_constructor_call);
> +  fprintf_unfiltered (file,
>                        "gdbarch_dump: gdbarch_skip_permanent_breakpoint_p()
> = %d\n",
>                        gdbarch_skip_permanent_breakpoint_p (gdbarch));
>    fprintf_unfiltered (file,
> @@ -2075,6 +2084,30 @@ set_gdbarch_skip_prologue (struct gdbarc
>  }
>
>  int
> +gdbarch_skip_main_constructor_call_p (struct gdbarch *gdbarch)
> +{
> +  gdb_assert (gdbarch != NULL);
> +  return gdbarch->skip_main_constructor_call != NULL;
> +}
> +
> +CORE_ADDR
> +gdbarch_skip_main_constructor_call (struct gdbarch *gdbarch, CORE_ADDR ip)
> +{
> +  gdb_assert (gdbarch != NULL);
> +  gdb_assert (gdbarch->skip_main_constructor_call != NULL);
> +  if (gdbarch_debug >= 2)
> +    fprintf_unfiltered (gdb_stdlog, "gdbarch_skip_main_constructor_call
> called\n");
> +  return gdbarch->skip_main_constructor_call (gdbarch, ip);
> +}
> +
> +void
> +set_gdbarch_skip_main_constructor_call (struct gdbarch *gdbarch,
> +                              gdbarch_skip_main_constructor_call_ftype
> skip_main_constructor_call)
> +{
> +  gdbarch->skip_main_constructor_call = skip_main_constructor_call;
> +}
> +
> +int
>  gdbarch_inner_than (struct gdbarch *gdbarch, CORE_ADDR lhs, CORE_ADDR rhs)
>  {
>    gdb_assert (gdbarch != NULL);
> Index: gdb/gdbarch.h
> ===================================================================
> RCS file: /cvs/src/src/gdb/gdbarch.h,v
> retrieving revision 1.377
> diff -u -p -r1.377 gdbarch.h
> --- gdb/gdbarch.h	29 Apr 2008 16:06:07 -0000	1.377
> +++ gdb/gdbarch.h	30 Apr 2008 13:01:33 -0000
> @@ -381,6 +381,12 @@ typedef CORE_ADDR (gdbarch_skip_prologue
>  extern CORE_ADDR gdbarch_skip_prologue (struct gdbarch *gdbarch, CORE_ADDR
> ip);
>  extern void set_gdbarch_skip_prologue (struct gdbarch *gdbarch,
> gdbarch_skip_prologue_ftype *skip_prologue);
>
> +extern int gdbarch_skip_main_constructor_call_p (struct gdbarch *gdbarch);
> +
> +typedef CORE_ADDR (gdbarch_skip_main_constructor_call_ftype) (struct
> gdbarch *gdbarch, CORE_ADDR ip);
> +extern CORE_ADDR gdbarch_skip_main_constructor_call (struct gdbarch
> *gdbarch, CORE_ADDR ip);
> +extern void set_gdbarch_skip_main_constructor_call (struct gdbarch
> *gdbarch, gdbarch_skip_main_constructor_call_ftype
> *skip_main_constructor_call);
> +
>  typedef int (gdbarch_inner_than_ftype) (CORE_ADDR lhs, CORE_ADDR rhs);
>  extern int gdbarch_inner_than (struct gdbarch *gdbarch, CORE_ADDR lhs,
> CORE_ADDR rhs);
>  extern void set_gdbarch_inner_than (struct gdbarch *gdbarch,
> gdbarch_inner_than_ftype *inner_than);
> Index: gdb/gdbarch.sh
> ===================================================================
> RCS file: /cvs/src/src/gdb/gdbarch.sh,v
> retrieving revision 1.463
> diff -u -p -r1.463 gdbarch.sh
> --- gdb/gdbarch.sh	29 Apr 2008 16:06:06 -0000	1.463
> +++ gdb/gdbarch.sh	30 Apr 2008 13:01:34 -0000
> @@ -483,6 +483,7 @@ M:CORE_ADDR:integer_to_address:struct ty
>  M:enum return_value_convention:return_value:struct type *functype, struct
> type *valtype, struct regcache *regcache, gdb_byte *readbuf, const gdb_byte
> *writebuf:functype, valtype, regcache, readbuf, writebuf
>
>  m:CORE_ADDR:skip_prologue:CORE_ADDR ip:ip:0:0
> +M:CORE_ADDR:skip_main_constructor_call:CORE_ADDR ip:ip
>  f:int:inner_than:CORE_ADDR lhs, CORE_ADDR rhs:lhs, rhs:0:0
>  m:const gdb_byte *:breakpoint_from_pc:CORE_ADDR *pcptr, int *lenptr:pcptr,
> lenptr::0:
>  M:CORE_ADDR:adjust_breakpoint_address:CORE_ADDR bpaddr:bpaddr
> Index: gdb/i386-cygwin-tdep.c
> ===================================================================
> RCS file: /cvs/src/src/gdb/i386-cygwin-tdep.c,v
> retrieving revision 1.16
> diff -u -p -r1.16 i386-cygwin-tdep.c
> --- gdb/i386-cygwin-tdep.c	1 Jan 2008 22:53:10 -0000	1.16
> +++ gdb/i386-cygwin-tdep.c	30 Apr 2008 13:01:34 -0000
> @@ -227,6 +227,8 @@ i386_cygwin_init_abi (struct gdbarch_inf
>
>    set_gdbarch_skip_trampoline_code (gdbarch,
> i386_cygwin_skip_trampoline_code);
>
> +  set_gdbarch_skip_main_constructor_call (gdbarch,
> i386_skip_main_constructor_call);
> +
>    tdep->struct_return = reg_struct_return;
>
>    tdep->gregset_reg_offset = i386_win32_gregset_reg_offset;
> Index: gdb/i386-tdep.c
> ===================================================================
> RCS file: /cvs/src/src/gdb/i386-tdep.c,v
> retrieving revision 1.255
> diff -u -p -r1.255 i386-tdep.c
> --- gdb/i386-tdep.c	25 Apr 2008 14:57:30 -0000	1.255
> +++ gdb/i386-tdep.c	30 Apr 2008 13:01:35 -0000
> @@ -941,6 +941,33 @@ i386_skip_prologue (struct gdbarch *gdba
>    return pc;
>  }
>
> +/* Check that the code pointed to by PC corresponds to a call to
> +   __main, skip it if so.  Return PC otherwise.  */
> +
> +CORE_ADDR
> +i386_skip_main_constructor_call (struct gdbarch *gdbarch, CORE_ADDR pc)
> +{
> +  gdb_byte op;
> +
> +  target_read_memory (pc, &op, 1);
> +  if (op == 0xe8)
> +    {
> +      gdb_byte buf[4];
> +      if (target_read_memory (pc + 1, buf, sizeof buf) == 0)
> + 	{
> + 	  CORE_ADDR call_dest = pc + 5 + extract_unsigned_integer (buf, 4);
> +
> + 	  struct minimal_symbol *s = lookup_minimal_symbol_by_pc
> (call_dest);
> + 	  if (s != NULL
> + 	      && SYMBOL_LINKAGE_NAME (s) != NULL
> + 	      && strcmp (SYMBOL_LINKAGE_NAME (s), "__main") == 0)
> + 	    pc += 5;
> + 	}
> +    }
> +
> +  return pc;
> +}
> +
>  /* This function is 64-bit safe.  */
>
>  static CORE_ADDR
> Index: gdb/i386-tdep.h
> ===================================================================
> RCS file: /cvs/src/src/gdb/i386-tdep.h,v
> retrieving revision 1.53
> diff -u -p -r1.53 i386-tdep.h
> --- gdb/i386-tdep.h	11 Mar 2008 05:21:38 -0000	1.53
> +++ gdb/i386-tdep.h	30 Apr 2008 13:01:35 -0000
> @@ -166,6 +166,7 @@ extern struct type *i386_sse_type (struc
>
>  /* Functions exported from i386-tdep.c.  */
>  extern CORE_ADDR i386_pe_skip_trampoline_code (CORE_ADDR pc, char *name);
> +extern CORE_ADDR i386_skip_main_constructor_call (struct gdbarch *gdbarch,
> CORE_ADDR pc);
>
>  /* Return the name of register REGNUM.  */
>  extern char const *i386_register_name (struct gdbarch * gdbarch, int
> regnum);
> Index: gdb/symtab.c
> ===================================================================
> RCS file: /cvs/src/src/gdb/symtab.c,v
> retrieving revision 1.177
> diff -u -p -r1.177 symtab.c
> --- gdb/symtab.c	19 Apr 2008 11:39:50 -0000	1.177
> +++ gdb/symtab.c	30 Apr 2008 13:01:36 -0000
> @@ -2572,6 +2572,21 @@ find_function_start_sal (struct symbol *
>        /* Recalculate the line number (might not be N+1).  */
>        sal = find_pc_sect_line (pc, SYMBOL_BFD_SECTION (sym), 0);
>      }
> +
> +  /* On targets with executable formats that don't have a concept of
> +     constructors (ELF with .init has, PE doesn't), gcc emits a call
> +     to `__main' in `main' between the prologue and before user
> +     code.  */
> +  if (funfirstline
> +      && gdbarch_skip_main_constructor_call_p (current_gdbarch)
> +      && SYMBOL_LINKAGE_NAME (sym)
> +      && strcmp (SYMBOL_LINKAGE_NAME (sym), "main") == 0)
> +    {
> +      pc = gdbarch_skip_main_constructor_call (current_gdbarch, pc);
> +      /* Recalculate the line number (might not be N+1).  */
> +      sal = find_pc_sect_line (pc, SYMBOL_BFD_SECTION (sym), 0);
> +    }
> +
>    sal.pc = pc;
>
>    return sal;

[PING2] : [RFC/RFA] PING: skip __main

[Pierre Muller]

Hi, is there any chance that someone will have
the time to review this patch?


Pierre Muller
Pascal language support maintainer for GDB



-----Message d'origine-----
De : gdb-patches-owner@sourceware.org
[mailto:gdb-patches-owner@sourceware.org] De la part de Pierre Muller
Envoyé : Friday, May 02, 2008 3:26 PM
À : 'Pedro Alves'; 'Daniel Jacobowitz'; gdb-patches@sourceware.org
Cc : 'Mark Kettenis'
Objet : [RFC/RFA] PING: skip __main

  Pedro submitted this patch end of January,
but the 6.8 release pushed it to a later date.
http://sourceware.org/ml/gdb-patches/2008-01/msg00665.html
  
Now that GDB 6.8 is out, 
I would really like to get this patch in.
Pedro, this is your work, but you seem to be
much more involved in other things lately,
thus, if you don't mind, I can try to push this through. 

  I updated the patch from the first message
referenced and just followed Daniel's advice in
http://sourceware.org/ml/gdb-patches/2008-01/msg00883.html
to change skip___main into skip_main_constructor.

  This __main call seems to be
a common feature for coff format, thus 
I suppose that it applies to other targets,
go32v2 djgpp is probably one of them.

For cygwin target, I get this:

                === gdb Summary ===
	
-# of expected passes           10733
-# of unexpected failures       560
+# of expected passes           10928
+# of unexpected failures       363
 # of expected failures         59
 # of unknown successes         2
-# of known failures            21
+# of known failures            23
 # of unresolved testcases      40
 # of untested testcases                14
 # of unsupported tests         23

  Almost 200 failures less...

  Daniel wanted to get a comment from Mark
in his last email in that thread, that is the reason
why I added Mark Kettenis to the list of recipients.

Does this patch look OK now?

  One point that should still be discussed is
if we should call main_name function
rather than hardcoded "main" in find_function_start_sal in symtab.c.

Pierre Muller
Pascal language support maintainer for GDB


2008-05-02  Pedro Alves  <pedro_alves@portugalmail.pt>
	Pierre Muller  <muller@ics.u-strasbg.fr>

	* gdbarch.sh (gdbarch_skip_main_constructor_call): New.
	* gdbarch.h, gdbarch.c: Regenerate.

	* i386-tdep.h (i386_skip_main_constructor_call): Declare.
	* i386-tdep.c (i386_skip_main_constructor_call): New.
	* i386-cygwin-tdep.c (i386_cygwin_init_abi): Register
	i386_skip_main_constructor_call as
gdbarch_skip_main_constructor_call 
	gdbarch callback.
	* symtab.c (find_function_start_sal): When pc points at the "main"
	function, call gdbarch_skip_main_constructor_call.

Index: gdb/gdbarch.c
===================================================================
RCS file: /cvs/src/src/gdb/gdbarch.c,v
retrieving revision 1.422
diff -u -p -r1.422 gdbarch.c
--- gdb/gdbarch.c	22 Apr 2008 11:03:41 -0000	1.422
+++ gdb/gdbarch.c	30 Apr 2008 13:01:33 -0000
@@ -184,6 +184,7 @@ struct gdbarch
   gdbarch_integer_to_address_ftype *integer_to_address;
   gdbarch_return_value_ftype *return_value;
   gdbarch_skip_prologue_ftype *skip_prologue;
+  gdbarch_skip_main_constructor_call_ftype *skip_main_constructor_call;
   gdbarch_inner_than_ftype *inner_than;
   gdbarch_breakpoint_from_pc_ftype *breakpoint_from_pc;
   gdbarch_adjust_breakpoint_address_ftype *adjust_breakpoint_address;
@@ -306,6 +307,7 @@ struct gdbarch startup_gdbarch =
   0,  /* integer_to_address */
   0,  /* return_value */
   0,  /* skip_prologue */
+  0,  /* skip_main_constructor_call */
   0,  /* inner_than */
   0,  /* breakpoint_from_pc */
   0,  /* adjust_breakpoint_address */
@@ -542,6 +544,7 @@ verify_gdbarch (struct gdbarch *gdbarch)
   /* Skip verify of return_value, has predicate */
   if (gdbarch->skip_prologue == 0)
     fprintf_unfiltered (log, "\n\tskip_prologue");
+  /* Skip verify of skip_main_constructor_call, has predicate */
   if (gdbarch->inner_than == 0)
     fprintf_unfiltered (log, "\n\tinner_than");
   if (gdbarch->breakpoint_from_pc == 0)
@@ -934,6 +937,12 @@ gdbarch_dump (struct gdbarch *gdbarch, s
                       "gdbarch_dump: single_step_through_delay =
<0x%lx>\n",
                       (long) gdbarch->single_step_through_delay);
   fprintf_unfiltered (file,
+                      "gdbarch_dump: gdbarch_skip_main_constructor_call_p()
= %d\n",
+                      gdbarch_skip_main_constructor_call_p (gdbarch));
+  fprintf_unfiltered (file,
+                      "gdbarch_dump: skip_main_constructor_call =
<0x%lx>\n",
+                      (long) gdbarch->skip_main_constructor_call);
+  fprintf_unfiltered (file,
                       "gdbarch_dump: gdbarch_skip_permanent_breakpoint_p()
= %d\n",
                       gdbarch_skip_permanent_breakpoint_p (gdbarch));
   fprintf_unfiltered (file,
@@ -2075,6 +2084,30 @@ set_gdbarch_skip_prologue (struct gdbarc
 }
 
 int
+gdbarch_skip_main_constructor_call_p (struct gdbarch *gdbarch)
+{
+  gdb_assert (gdbarch != NULL);
+  return gdbarch->skip_main_constructor_call != NULL;
+}
+
+CORE_ADDR
+gdbarch_skip_main_constructor_call (struct gdbarch *gdbarch, CORE_ADDR ip)
+{
+  gdb_assert (gdbarch != NULL);
+  gdb_assert (gdbarch->skip_main_constructor_call != NULL);
+  if (gdbarch_debug >= 2)
+    fprintf_unfiltered (gdb_stdlog, "gdbarch_skip_main_constructor_call
called\n");
+  return gdbarch->skip_main_constructor_call (gdbarch, ip);
+}
+
+void
+set_gdbarch_skip_main_constructor_call (struct gdbarch *gdbarch,
+                              gdbarch_skip_main_constructor_call_ftype
skip_main_constructor_call)
+{
+  gdbarch->skip_main_constructor_call = skip_main_constructor_call;
+}
+
+int
 gdbarch_inner_than (struct gdbarch *gdbarch, CORE_ADDR lhs, CORE_ADDR rhs)
 {
   gdb_assert (gdbarch != NULL);
Index: gdb/gdbarch.h
===================================================================
RCS file: /cvs/src/src/gdb/gdbarch.h,v
retrieving revision 1.377
diff -u -p -r1.377 gdbarch.h
--- gdb/gdbarch.h	29 Apr 2008 16:06:07 -0000	1.377
+++ gdb/gdbarch.h	30 Apr 2008 13:01:33 -0000
@@ -381,6 +381,12 @@ typedef CORE_ADDR (gdbarch_skip_prologue
 extern CORE_ADDR gdbarch_skip_prologue (struct gdbarch *gdbarch, CORE_ADDR
ip);
 extern void set_gdbarch_skip_prologue (struct gdbarch *gdbarch,
gdbarch_skip_prologue_ftype *skip_prologue);
 
+extern int gdbarch_skip_main_constructor_call_p (struct gdbarch *gdbarch);
+
+typedef CORE_ADDR (gdbarch_skip_main_constructor_call_ftype) (struct
gdbarch *gdbarch, CORE_ADDR ip);
+extern CORE_ADDR gdbarch_skip_main_constructor_call (struct gdbarch
*gdbarch, CORE_ADDR ip);
+extern void set_gdbarch_skip_main_constructor_call (struct gdbarch
*gdbarch, gdbarch_skip_main_constructor_call_ftype
*skip_main_constructor_call);
+
 typedef int (gdbarch_inner_than_ftype) (CORE_ADDR lhs, CORE_ADDR rhs);
 extern int gdbarch_inner_than (struct gdbarch *gdbarch, CORE_ADDR lhs,
CORE_ADDR rhs);
 extern void set_gdbarch_inner_than (struct gdbarch *gdbarch,
gdbarch_inner_than_ftype *inner_than);
Index: gdb/gdbarch.sh
===================================================================
RCS file: /cvs/src/src/gdb/gdbarch.sh,v
retrieving revision 1.463
diff -u -p -r1.463 gdbarch.sh
--- gdb/gdbarch.sh	29 Apr 2008 16:06:06 -0000	1.463
+++ gdb/gdbarch.sh	30 Apr 2008 13:01:34 -0000
@@ -483,6 +483,7 @@ M:CORE_ADDR:integer_to_address:struct ty
 M:enum return_value_convention:return_value:struct type *functype, struct
type *valtype, struct regcache *regcache, gdb_byte *readbuf, const gdb_byte
*writebuf:functype, valtype, regcache, readbuf, writebuf
 
 m:CORE_ADDR:skip_prologue:CORE_ADDR ip:ip:0:0
+M:CORE_ADDR:skip_main_constructor_call:CORE_ADDR ip:ip
 f:int:inner_than:CORE_ADDR lhs, CORE_ADDR rhs:lhs, rhs:0:0
 m:const gdb_byte *:breakpoint_from_pc:CORE_ADDR *pcptr, int *lenptr:pcptr,
lenptr::0:
 M:CORE_ADDR:adjust_breakpoint_address:CORE_ADDR bpaddr:bpaddr
Index: gdb/i386-cygwin-tdep.c
===================================================================
RCS file: /cvs/src/src/gdb/i386-cygwin-tdep.c,v
retrieving revision 1.16
diff -u -p -r1.16 i386-cygwin-tdep.c
--- gdb/i386-cygwin-tdep.c	1 Jan 2008 22:53:10 -0000	1.16
+++ gdb/i386-cygwin-tdep.c	30 Apr 2008 13:01:34 -0000
@@ -227,6 +227,8 @@ i386_cygwin_init_abi (struct gdbarch_inf
 
   set_gdbarch_skip_trampoline_code (gdbarch,
i386_cygwin_skip_trampoline_code);
 
+  set_gdbarch_skip_main_constructor_call (gdbarch,
i386_skip_main_constructor_call);
+
   tdep->struct_return = reg_struct_return;
 
   tdep->gregset_reg_offset = i386_win32_gregset_reg_offset;
Index: gdb/i386-tdep.c
===================================================================
RCS file: /cvs/src/src/gdb/i386-tdep.c,v
retrieving revision 1.255
diff -u -p -r1.255 i386-tdep.c
--- gdb/i386-tdep.c	25 Apr 2008 14:57:30 -0000	1.255
+++ gdb/i386-tdep.c	30 Apr 2008 13:01:35 -0000
@@ -941,6 +941,33 @@ i386_skip_prologue (struct gdbarch *gdba
   return pc;
 }
 
+/* Check that the code pointed to by PC corresponds to a call to
+   __main, skip it if so.  Return PC otherwise.  */
+
+CORE_ADDR
+i386_skip_main_constructor_call (struct gdbarch *gdbarch, CORE_ADDR pc)
+{
+  gdb_byte op;
+
+  target_read_memory (pc, &op, 1);
+  if (op == 0xe8)
+    {
+      gdb_byte buf[4];
+      if (target_read_memory (pc + 1, buf, sizeof buf) == 0)
+ 	{
+ 	  CORE_ADDR call_dest = pc + 5 + extract_unsigned_integer (buf, 4);
+
+ 	  struct minimal_symbol *s = lookup_minimal_symbol_by_pc
(call_dest);
+ 	  if (s != NULL
+ 	      && SYMBOL_LINKAGE_NAME (s) != NULL
+ 	      && strcmp (SYMBOL_LINKAGE_NAME (s), "__main") == 0)
+ 	    pc += 5;
+ 	}
+    }
+
+  return pc;
+}
+
 /* This function is 64-bit safe.  */
 
 static CORE_ADDR
Index: gdb/i386-tdep.h
===================================================================
RCS file: /cvs/src/src/gdb/i386-tdep.h,v
retrieving revision 1.53
diff -u -p -r1.53 i386-tdep.h
--- gdb/i386-tdep.h	11 Mar 2008 05:21:38 -0000	1.53
+++ gdb/i386-tdep.h	30 Apr 2008 13:01:35 -0000
@@ -166,6 +166,7 @@ extern struct type *i386_sse_type (struc
 
 /* Functions exported from i386-tdep.c.  */
 extern CORE_ADDR i386_pe_skip_trampoline_code (CORE_ADDR pc, char *name);
+extern CORE_ADDR i386_skip_main_constructor_call (struct gdbarch *gdbarch,
CORE_ADDR pc);
 
 /* Return the name of register REGNUM.  */
 extern char const *i386_register_name (struct gdbarch * gdbarch, int
regnum);
Index: gdb/symtab.c
===================================================================
RCS file: /cvs/src/src/gdb/symtab.c,v
retrieving revision 1.177
diff -u -p -r1.177 symtab.c
--- gdb/symtab.c	19 Apr 2008 11:39:50 -0000	1.177
+++ gdb/symtab.c	30 Apr 2008 13:01:36 -0000
@@ -2572,6 +2572,21 @@ find_function_start_sal (struct symbol *
       /* Recalculate the line number (might not be N+1).  */
       sal = find_pc_sect_line (pc, SYMBOL_BFD_SECTION (sym), 0);
     }
+
+  /* On targets with executable formats that don't have a concept of
+     constructors (ELF with .init has, PE doesn't), gcc emits a call
+     to `__main' in `main' between the prologue and before user
+     code.  */
+  if (funfirstline
+      && gdbarch_skip_main_constructor_call_p (current_gdbarch)
+      && SYMBOL_LINKAGE_NAME (sym)
+      && strcmp (SYMBOL_LINKAGE_NAME (sym), "main") == 0)
+    {
+      pc = gdbarch_skip_main_constructor_call (current_gdbarch, pc);
+      /* Recalculate the line number (might not be N+1).  */
+      sal = find_pc_sect_line (pc, SYMBOL_BFD_SECTION (sym), 0);
+    }
+
   sal.pc = pc;
 
   return sal;

Re: [PING2] : [RFC/RFA] PING: skip __main

[Mark Kettenis]

> From: "Pierre Muller" <muller@ics.u-strasbg.fr>
> Date: Fri, 30 May 2008 13:18:20 +0200
> 
> Hi, is there any chance that someone will have
> the time to review this patch?

I'm afraid I think that skip_main_constructor_call is too long as a name.

I'd also appreciate it if you could seperate local variable
declarations from the stations that follow by a blank line.  Otherwise
this looks ok to me.

> -----Message d'origine-----
> De : gdb-patches-owner@sourceware.org
> [mailto:gdb-patches-owner@sourceware.org] De la part de Pierre Muller
> Envoyé : Friday, May 02, 2008 3:26 PM
> À : 'Pedro Alves'; 'Daniel Jacobowitz'; gdb-patches@sourceware.org
> Cc : 'Mark Kettenis'
> Objet : [RFC/RFA] PING: skip __main
> 
>   Pedro submitted this patch end of January,
> but the 6.8 release pushed it to a later date.
> http://sourceware.org/ml/gdb-patches/2008-01/msg00665.html
>   
> Now that GDB 6.8 is out, 
> I would really like to get this patch in.
> Pedro, this is your work, but you seem to be
> much more involved in other things lately,
> thus, if you don't mind, I can try to push this through. 
> 
>   I updated the patch from the first message
> referenced and just followed Daniel's advice in
> http://sourceware.org/ml/gdb-patches/2008-01/msg00883.html
> to change skip___main into skip_main_constructor.
> 
>   This __main call seems to be
> a common feature for coff format, thus 
> I suppose that it applies to other targets,
> go32v2 djgpp is probably one of them.
> 
> For cygwin target, I get this:
> 
>                 === gdb Summary ===
> 	
> -# of expected passes           10733
> -# of unexpected failures       560
> +# of expected passes           10928
> +# of unexpected failures       363
>  # of expected failures         59
>  # of unknown successes         2
> -# of known failures            21
> +# of known failures            23
>  # of unresolved testcases      40
>  # of untested testcases                14
>  # of unsupported tests         23
> 
>   Almost 200 failures less...
> 
>   Daniel wanted to get a comment from Mark
> in his last email in that thread, that is the reason
> why I added Mark Kettenis to the list of recipients.
> 
> Does this patch look OK now?
> 
>   One point that should still be discussed is
> if we should call main_name function
> rather than hardcoded "main" in find_function_start_sal in symtab.c.
> 
> Pierre Muller
> Pascal language support maintainer for GDB
> 
> 
> 2008-05-02  Pedro Alves  <pedro_alves@portugalmail.pt>
> 	Pierre Muller  <muller@ics.u-strasbg.fr>
> 
> 	* gdbarch.sh (gdbarch_skip_main_constructor_call): New.
> 	* gdbarch.h, gdbarch.c: Regenerate.
> 
> 	* i386-tdep.h (i386_skip_main_constructor_call): Declare.
> 	* i386-tdep.c (i386_skip_main_constructor_call): New.
> 	* i386-cygwin-tdep.c (i386_cygwin_init_abi): Register
> 	i386_skip_main_constructor_call as
> gdbarch_skip_main_constructor_call 
> 	gdbarch callback.
> 	* symtab.c (find_function_start_sal): When pc points at the "main"
> 	function, call gdbarch_skip_main_constructor_call.
> 
> Index: gdb/gdbarch.c
> ===================================================================
> RCS file: /cvs/src/src/gdb/gdbarch.c,v
> retrieving revision 1.422
> diff -u -p -r1.422 gdbarch.c
> --- gdb/gdbarch.c	22 Apr 2008 11:03:41 -0000	1.422
> +++ gdb/gdbarch.c	30 Apr 2008 13:01:33 -0000
> @@ -184,6 +184,7 @@ struct gdbarch
>    gdbarch_integer_to_address_ftype *integer_to_address;
>    gdbarch_return_value_ftype *return_value;
>    gdbarch_skip_prologue_ftype *skip_prologue;
> +  gdbarch_skip_main_constructor_call_ftype *skip_main_constructor_call;
>    gdbarch_inner_than_ftype *inner_than;
>    gdbarch_breakpoint_from_pc_ftype *breakpoint_from_pc;
>    gdbarch_adjust_breakpoint_address_ftype *adjust_breakpoint_address;
> @@ -306,6 +307,7 @@ struct gdbarch startup_gdbarch =
>    0,  /* integer_to_address */
>    0,  /* return_value */
>    0,  /* skip_prologue */
> +  0,  /* skip_main_constructor_call */
>    0,  /* inner_than */
>    0,  /* breakpoint_from_pc */
>    0,  /* adjust_breakpoint_address */
> @@ -542,6 +544,7 @@ verify_gdbarch (struct gdbarch *gdbarch)
>    /* Skip verify of return_value, has predicate */
>    if (gdbarch->skip_prologue == 0)
>      fprintf_unfiltered (log, "\n\tskip_prologue");
> +  /* Skip verify of skip_main_constructor_call, has predicate */
>    if (gdbarch->inner_than == 0)
>      fprintf_unfiltered (log, "\n\tinner_than");
>    if (gdbarch->breakpoint_from_pc == 0)
> @@ -934,6 +937,12 @@ gdbarch_dump (struct gdbarch *gdbarch, s
>                        "gdbarch_dump: single_step_through_delay =
> <0x%lx>\n",
>                        (long) gdbarch->single_step_through_delay);
>    fprintf_unfiltered (file,
> +                      "gdbarch_dump: gdbarch_skip_main_constructor_call_p()
> = %d\n",
> +                      gdbarch_skip_main_constructor_call_p (gdbarch));
> +  fprintf_unfiltered (file,
> +                      "gdbarch_dump: skip_main_constructor_call =
> <0x%lx>\n",
> +                      (long) gdbarch->skip_main_constructor_call);
> +  fprintf_unfiltered (file,
>                        "gdbarch_dump: gdbarch_skip_permanent_breakpoint_p()
> = %d\n",
>                        gdbarch_skip_permanent_breakpoint_p (gdbarch));
>    fprintf_unfiltered (file,
> @@ -2075,6 +2084,30 @@ set_gdbarch_skip_prologue (struct gdbarc
>  }
>  
>  int
> +gdbarch_skip_main_constructor_call_p (struct gdbarch *gdbarch)
> +{
> +  gdb_assert (gdbarch != NULL);
> +  return gdbarch->skip_main_constructor_call != NULL;
> +}
> +
> +CORE_ADDR
> +gdbarch_skip_main_constructor_call (struct gdbarch *gdbarch, CORE_ADDR ip)
> +{
> +  gdb_assert (gdbarch != NULL);
> +  gdb_assert (gdbarch->skip_main_constructor_call != NULL);
> +  if (gdbarch_debug >= 2)
> +    fprintf_unfiltered (gdb_stdlog, "gdbarch_skip_main_constructor_call
> called\n");
> +  return gdbarch->skip_main_constructor_call (gdbarch, ip);
> +}
> +
> +void
> +set_gdbarch_skip_main_constructor_call (struct gdbarch *gdbarch,
> +                              gdbarch_skip_main_constructor_call_ftype
> skip_main_constructor_call)
> +{
> +  gdbarch->skip_main_constructor_call = skip_main_constructor_call;
> +}
> +
> +int
>  gdbarch_inner_than (struct gdbarch *gdbarch, CORE_ADDR lhs, CORE_ADDR rhs)
>  {
>    gdb_assert (gdbarch != NULL);
> Index: gdb/gdbarch.h
> ===================================================================
> RCS file: /cvs/src/src/gdb/gdbarch.h,v
> retrieving revision 1.377
> diff -u -p -r1.377 gdbarch.h
> --- gdb/gdbarch.h	29 Apr 2008 16:06:07 -0000	1.377
> +++ gdb/gdbarch.h	30 Apr 2008 13:01:33 -0000
> @@ -381,6 +381,12 @@ typedef CORE_ADDR (gdbarch_skip_prologue
>  extern CORE_ADDR gdbarch_skip_prologue (struct gdbarch *gdbarch, CORE_ADDR
> ip);
>  extern void set_gdbarch_skip_prologue (struct gdbarch *gdbarch,
> gdbarch_skip_prologue_ftype *skip_prologue);
>  
> +extern int gdbarch_skip_main_constructor_call_p (struct gdbarch *gdbarch);
> +
> +typedef CORE_ADDR (gdbarch_skip_main_constructor_call_ftype) (struct
> gdbarch *gdbarch, CORE_ADDR ip);
> +extern CORE_ADDR gdbarch_skip_main_constructor_call (struct gdbarch
> *gdbarch, CORE_ADDR ip);
> +extern void set_gdbarch_skip_main_constructor_call (struct gdbarch
> *gdbarch, gdbarch_skip_main_constructor_call_ftype
> *skip_main_constructor_call);
> +
>  typedef int (gdbarch_inner_than_ftype) (CORE_ADDR lhs, CORE_ADDR rhs);
>  extern int gdbarch_inner_than (struct gdbarch *gdbarch, CORE_ADDR lhs,
> CORE_ADDR rhs);
>  extern void set_gdbarch_inner_than (struct gdbarch *gdbarch,
> gdbarch_inner_than_ftype *inner_than);
> Index: gdb/gdbarch.sh
> ===================================================================
> RCS file: /cvs/src/src/gdb/gdbarch.sh,v
> retrieving revision 1.463
> diff -u -p -r1.463 gdbarch.sh
> --- gdb/gdbarch.sh	29 Apr 2008 16:06:06 -0000	1.463
> +++ gdb/gdbarch.sh	30 Apr 2008 13:01:34 -0000
> @@ -483,6 +483,7 @@ M:CORE_ADDR:integer_to_address:struct ty
>  M:enum return_value_convention:return_value:struct type *functype, struct
> type *valtype, struct regcache *regcache, gdb_byte *readbuf, const gdb_byte
> *writebuf:functype, valtype, regcache, readbuf, writebuf
>  
>  m:CORE_ADDR:skip_prologue:CORE_ADDR ip:ip:0:0
> +M:CORE_ADDR:skip_main_constructor_call:CORE_ADDR ip:ip
>  f:int:inner_than:CORE_ADDR lhs, CORE_ADDR rhs:lhs, rhs:0:0
>  m:const gdb_byte *:breakpoint_from_pc:CORE_ADDR *pcptr, int *lenptr:pcptr,
> lenptr::0:
>  M:CORE_ADDR:adjust_breakpoint_address:CORE_ADDR bpaddr:bpaddr
> Index: gdb/i386-cygwin-tdep.c
> ===================================================================
> RCS file: /cvs/src/src/gdb/i386-cygwin-tdep.c,v
> retrieving revision 1.16
> diff -u -p -r1.16 i386-cygwin-tdep.c
> --- gdb/i386-cygwin-tdep.c	1 Jan 2008 22:53:10 -0000	1.16
> +++ gdb/i386-cygwin-tdep.c	30 Apr 2008 13:01:34 -0000
> @@ -227,6 +227,8 @@ i386_cygwin_init_abi (struct gdbarch_inf
>  
>    set_gdbarch_skip_trampoline_code (gdbarch,
> i386_cygwin_skip_trampoline_code);
>  
> +  set_gdbarch_skip_main_constructor_call (gdbarch,
> i386_skip_main_constructor_call);
> +
>    tdep->struct_return = reg_struct_return;
>  
>    tdep->gregset_reg_offset = i386_win32_gregset_reg_offset;
> Index: gdb/i386-tdep.c
> ===================================================================
> RCS file: /cvs/src/src/gdb/i386-tdep.c,v
> retrieving revision 1.255
> diff -u -p -r1.255 i386-tdep.c
> --- gdb/i386-tdep.c	25 Apr 2008 14:57:30 -0000	1.255
> +++ gdb/i386-tdep.c	30 Apr 2008 13:01:35 -0000
> @@ -941,6 +941,33 @@ i386_skip_prologue (struct gdbarch *gdba
>    return pc;
>  }
>  
> +/* Check that the code pointed to by PC corresponds to a call to
> +   __main, skip it if so.  Return PC otherwise.  */
> +
> +CORE_ADDR
> +i386_skip_main_constructor_call (struct gdbarch *gdbarch, CORE_ADDR pc)
> +{
> +  gdb_byte op;
> +
> +  target_read_memory (pc, &op, 1);
> +  if (op == 0xe8)
> +    {
> +      gdb_byte buf[4];
> +      if (target_read_memory (pc + 1, buf, sizeof buf) == 0)
> + 	{
> + 	  CORE_ADDR call_dest = pc + 5 + extract_unsigned_integer (buf, 4);
> +
> + 	  struct minimal_symbol *s = lookup_minimal_symbol_by_pc
> (call_dest);
> + 	  if (s != NULL
> + 	      && SYMBOL_LINKAGE_NAME (s) != NULL
> + 	      && strcmp (SYMBOL_LINKAGE_NAME (s), "__main") == 0)
> + 	    pc += 5;
> + 	}
> +    }
> +
> +  return pc;
> +}
> +
>  /* This function is 64-bit safe.  */
>  
>  static CORE_ADDR
> Index: gdb/i386-tdep.h
> ===================================================================
> RCS file: /cvs/src/src/gdb/i386-tdep.h,v
> retrieving revision 1.53
> diff -u -p -r1.53 i386-tdep.h
> --- gdb/i386-tdep.h	11 Mar 2008 05:21:38 -0000	1.53
> +++ gdb/i386-tdep.h	30 Apr 2008 13:01:35 -0000
> @@ -166,6 +166,7 @@ extern struct type *i386_sse_type (struc
>  
>  /* Functions exported from i386-tdep.c.  */
>  extern CORE_ADDR i386_pe_skip_trampoline_code (CORE_ADDR pc, char *name);
> +extern CORE_ADDR i386_skip_main_constructor_call (struct gdbarch *gdbarch,
> CORE_ADDR pc);
>  
>  /* Return the name of register REGNUM.  */
>  extern char const *i386_register_name (struct gdbarch * gdbarch, int
> regnum);
> Index: gdb/symtab.c
> ===================================================================
> RCS file: /cvs/src/src/gdb/symtab.c,v
> retrieving revision 1.177
> diff -u -p -r1.177 symtab.c
> --- gdb/symtab.c	19 Apr 2008 11:39:50 -0000	1.177
> +++ gdb/symtab.c	30 Apr 2008 13:01:36 -0000
> @@ -2572,6 +2572,21 @@ find_function_start_sal (struct symbol *
>        /* Recalculate the line number (might not be N+1).  */
>        sal = find_pc_sect_line (pc, SYMBOL_BFD_SECTION (sym), 0);
>      }
> +
> +  /* On targets with executable formats that don't have a concept of
> +     constructors (ELF with .init has, PE doesn't), gcc emits a call
> +     to `__main' in `main' between the prologue and before user
> +     code.  */
> +  if (funfirstline
> +      && gdbarch_skip_main_constructor_call_p (current_gdbarch)
> +      && SYMBOL_LINKAGE_NAME (sym)
> +      && strcmp (SYMBOL_LINKAGE_NAME (sym), "main") == 0)
> +    {
> +      pc = gdbarch_skip_main_constructor_call (current_gdbarch, pc);
> +      /* Recalculate the line number (might not be N+1).  */
> +      sal = find_pc_sect_line (pc, SYMBOL_BFD_SECTION (sym), 0);
> +    }
> +
>    sal.pc = pc;
>  
>    return sal;
> 
> 
> 
> 
> 
> 

RE: [PING2] : [RFC/RFA] PING: skip __main

[Pierre Muller]

Thanks for the fast reply.

> I'm afraid I think that skip_main_constructor_call is too long as a name.

  I am perfectly willing to use something shorter, 
the only problem is to find something that would still be 
of clear meaning.

   maybe
  skip_main_prologue
would be better?

> I'd also appreciate it if you could seperate local variable
> declarations from the stations that follow by a blank line.  Otherwise
> this looks ok to me.

  Does this apply to both uninitialized and initialized variables?

Would the code hereafter be correct, or did I add too many empty lines?


+/* Check that the code pointed to by PC corresponds to a call to
+   __main, skip it if so.  Return PC otherwise.  */
+
+CORE_ADDR
+i386_skip_main_constructor_call (struct gdbarch *gdbarch, CORE_ADDR pc)
+{
+  gdb_byte op;
+
+  target_read_memory (pc, &op, 1);
+  if (op == 0xe8)
+    {
+      gdb_byte buf[4];
+
+      if (target_read_memory (pc + 1, buf, sizeof buf) == 0)
+       {
+         CORE_ADDR call_dest = pc + 5 + extract_unsigned_integer (buf, 4);
+
+         struct minimal_symbol *s = lookup_minimal_symbol_by_pc
(call_dest);
+
+         if (s != NULL
+             && SYMBOL_LINKAGE_NAME (s) != NULL
+             && strcmp (SYMBOL_LINKAGE_NAME (s), "__main") == 0)
+           pc += 5;
+       }
+    }
+
+  return pc;
+}
+

Thanks again for the reply.


Pierre Muller
Pascal language support maintainer for GDB

RE: [PING2] : [RFC/RFA] PING: skip __main

[Pierre Muller]

+  target_read_memory (pc, &op, 1);
+  if (op == 0xe8)
+    {
+      gdb_byte buf[4];
+
+      if (target_read_memory (pc + 1, buf, sizeof buf) == 0)
+       {
+         CORE_ADDR call_dest = pc + 5 + extract_unsigned_integer (buf, 4);
  Just one small question about this line:
I fear that this line is incorrect,
shouldn't we use extract_signed_integer here?

  The address is relative to pc+5 meaning that
if main is at a higher address than __main,
buf will contain a negative 4byte integer.

  This will usually not matter (if overflows are allowed), 
unless CORE_ADDR is 8-byte long, 
which would then give a value above 0xffffffff.

  This can happen for cross configured gdb, but also if gdb is configured
with --enable-64bit-bfd.

  I don't even know if these result would be always correct
(assuming for instance that main is below __main, but that
__main resides above 0x80000000 boundary for example).

  Maybe the only correct way is to force the operation to be performed
with 4-byte integers and take the result, even if overflow
occurred.


Pierre Muller
Pascal language support maintainer for GDB



  

Re: [PING2] : [RFC/RFA] PING: skip __main

[Mark Kettenis]

> From: "Pierre Muller" <muller@ics.u-strasbg.fr>
> Date: Fri, 30 May 2008 16:46:28 +0200
> 
> +  target_read_memory (pc, &op, 1);
> +  if (op == 0xe8)
> +    {
> +      gdb_byte buf[4];
> +
> +      if (target_read_memory (pc + 1, buf, sizeof buf) == 0)
> +       {
> +         CORE_ADDR call_dest = pc + 5 + extract_unsigned_integer (buf, 4);
>   Just one small question about this line:
> I fear that this line is incorrect,
> shouldn't we use extract_signed_integer here?
> 
>   The address is relative to pc+5 meaning that
> if main is at a higher address than __main,
> buf will contain a negative 4byte integer.

You're absolutely right.  Can you make that change and retest?

RE: [PING2] : [RFC/RFA] PING: skip __main

[Pierre Muller]

I restarted a testsuite, but I 
won't expect any change because 
main seems to be at a lower address than __main
in the few executables I tested, meaning that
the value in buf will be positive anyhow.


Pierre Muller
Pascal language support maintainer for GDB




-----Message d'origine-----
De : gdb-patches-owner@sourceware.org
[mailto:gdb-patches-owner@sourceware.org] De la part de Mark Kettenis
Envoyé : Friday, May 30, 2008 4:58 PM
À : muller@ics.u-strasbg.fr
Cc : pedro@codesourcery.com; drow@false.org; gdb-patches@sourceware.org
Objet : Re: [PING2] : [RFC/RFA] PING: skip __main

> From: "Pierre Muller" <muller@ics.u-strasbg.fr>
> Date: Fri, 30 May 2008 16:46:28 +0200
> 
> +  target_read_memory (pc, &op, 1);
> +  if (op == 0xe8)
> +    {
> +      gdb_byte buf[4];
> +
> +      if (target_read_memory (pc + 1, buf, sizeof buf) == 0)
> +       {
> +         CORE_ADDR call_dest = pc + 5 + extract_unsigned_integer (buf,
4);
>   Just one small question about this line:
> I fear that this line is incorrect,
> shouldn't we use extract_signed_integer here?
> 
>   The address is relative to pc+5 meaning that
> if main is at a higher address than __main,
> buf will contain a negative 4byte integer.

You're absolutely right.  Can you make that change and retest?

Re: [PING2] : [RFC/RFA] PING: skip __main

[Daniel Jacobowitz]

On Fri, May 30, 2008 at 03:06:51PM +0200, Pierre Muller wrote:
> Thanks for the fast reply.
> 
> > I'm afraid I think that skip_main_constructor_call is too long as a name.
> 
>   I am perfectly willing to use something shorter, 
> the only problem is to find something that would still be 
> of clear meaning.
> 
>    maybe
>   skip_main_prologue
> would be better?

Just skip_main_constructor?

-- 
Daniel Jacobowitz
CodeSourcery

Re: [PING2] : [RFC/RFA] PING: skip __main

[Mark Kettenis]

> From: "Pierre Muller" <muller@ics.u-strasbg.fr>
> Date: Fri, 30 May 2008 15:06:51 +0200
> 
> > I'm afraid I think that skip_main_constructor_call is too long as a name.
> 
>   I am perfectly willing to use something shorter, 
> the only problem is to find something that would still be 
> of clear meaning.
> 
>    maybe
>   skip_main_prologue
> would be better?

Actually, I think it is.  The functionality can be used for other
prologue skipping that's not really related to constructors as well.

> > I'd also appreciate it if you could seperate local variable
> > declarations from the stations that follow by a blank line.  Otherwise
> > this looks ok to me.
> 
>   Does this apply to both uninitialized and initialized variables?
> 
> Would the code hereafter be correct, or did I add too many empty lines?

Please remove the blank line before the "struct minimal_symbol *s"
declaration.

> +/* Check that the code pointed to by PC corresponds to a call to
> +   __main, skip it if so.  Return PC otherwise.  */
> +
> +CORE_ADDR
> +i386_skip_main_constructor_call (struct gdbarch *gdbarch, CORE_ADDR pc)
> +{
> +  gdb_byte op;
> +
> +  target_read_memory (pc, &op, 1);
> +  if (op == 0xe8)
> +    {
> +      gdb_byte buf[4];
> +
> +      if (target_read_memory (pc + 1, buf, sizeof buf) == 0)
> +       {
> +         CORE_ADDR call_dest = pc + 5 + extract_unsigned_integer (buf, 4);
> +
> +         struct minimal_symbol *s = lookup_minimal_symbol_by_pc
> (call_dest);
> +
> +         if (s != NULL
> +             && SYMBOL_LINKAGE_NAME (s) != NULL
> +             && strcmp (SYMBOL_LINKAGE_NAME (s), "__main") == 0)
> +           pc += 5;
> +       }
> +    }
> +
> +  return pc;
> +}
> +

Re: [PING2] : [RFC/RFA] PING: skip __main

[Daniel Jacobowitz]

On Sat, May 31, 2008 at 08:46:25AM +0200, Pierre Muller wrote:
>   Should I use the 32bit typecast strategy:
> 
> unsigned int32 pc_after = pc + 5; 
> unsigned int32 depl = extract_unsigned_integer (buf, 4);
>    
> unsigned int32 dest = pc_after - depl;
> pc = dest;
> 
> I don't know the checks done in C,
> are there any overflow checks in code like this,
> or is it safe to assume that it will work,
> even on machine that perform their operations
> on more than 32 bits?

If you use uint32_t, then you are guaranteed wrapping at 32 bits.
So that is probably the best thing to do.  See the patch I just posted
to make stdint.h available everywhere.

-- 
Daniel Jacobowitz
CodeSourcery

Re: [PING2] : [RFC/RFA] PING: skip __main

[Mark Kettenis]

> Date: Thu, 5 Jun 2008 16:26:40 -0400
> From: Daniel Jacobowitz <drow@false.org>
> 
> On Sat, May 31, 2008 at 08:46:25AM +0200, Pierre Muller wrote:
> >   Should I use the 32bit typecast strategy:
> > 
> > unsigned int32 pc_after = pc + 5; 
> > unsigned int32 depl = extract_unsigned_integer (buf, 4);
> >    
> > unsigned int32 dest = pc_after - depl;
> > pc = dest;
> > 
> > I don't know the checks done in C,
> > are there any overflow checks in code like this,
> > or is it safe to assume that it will work,
> > even on machine that perform their operations
> > on more than 32 bits?
> 
> If you use uint32_t, then you are guaranteed wrapping at 32 bits.
> So that is probably the best thing to do.  See the patch I just posted
> to make stdint.h available everywhere.

The diff should use extract_signed_integer().

I'm not sure if the wrapping is actually necessary.  Does the
assembler/linker actually depend on it?  Or does it error out on such
large jumps?  If so, alternatively you could explicity truncate the
generated address to 32 bits using & 0xffffffffU.

RE: [PING2] : [RFC/RFA] PING: skip __main

[Pierre Muller]

 I am not sure I understand your question about
assembler or linker, but
for sure if we call lookup_minimal_symbol_by_pc with
a CORE_ADDR outside the 0 to 0xffffffff range,
it might fail to find the right location
(although I must admit that I didn't try to look inside that
code to see if there is a bit truncation there)

  Anyhow, using extract_signed_integer followed by
a "& 0xffffffffU", we should be on the safe side.
  I will rerun a test a resubmit a modified patch after.



Pierre Muller
Pascal language support maintainer for GDB


  

-----Message d'origine-----
De : Mark Kettenis [mailto:mark.kettenis@xs4all.nl] 
Envoyé : Thursday, June 05, 2008 10:42 PM
À : drow@false.org
Cc : muller@ics.u-strasbg.fr; pedro@codesourcery.com;
gdb-patches@sourceware.org
Objet : Re: [PING2] : [RFC/RFA] PING: skip __main

> Date: Thu, 5 Jun 2008 16:26:40 -0400
> From: Daniel Jacobowitz <drow@false.org>
> 
> On Sat, May 31, 2008 at 08:46:25AM +0200, Pierre Muller wrote:
> >   Should I use the 32bit typecast strategy:
> > 
> > unsigned int32 pc_after = pc + 5; 
> > unsigned int32 depl = extract_unsigned_integer (buf, 4);
> >    
> > unsigned int32 dest = pc_after - depl;
> > pc = dest;
> > 
> > I don't know the checks done in C,
> > are there any overflow checks in code like this,
> > or is it safe to assume that it will work,
> > even on machine that perform their operations
> > on more than 32 bits?
> 
> If you use uint32_t, then you are guaranteed wrapping at 32 bits.
> So that is probably the best thing to do.  See the patch I just posted
> to make stdint.h available everywhere.

The diff should use extract_signed_integer().

I'm not sure if the wrapping is actually necessary.  Does the
assembler/linker actually depend on it?  Or does it error out on such
large jumps?  If so, alternatively you could explicity truncate the
generated address to 32 bits using & 0xffffffffU.

[RFA] New skip __main version

[Pierre Muller]

Here is a new version of this patch:
it should fix the concern I raised about
potential wrapping problems if
CORE_ADDR is 64 bit wide.
  I am still unable to test the 64-bit as
the --enable-target=all --enable-64bit-bfd 
is still broken.
  A patch has been submitted by Ulrich Weigand
http://sourceware.org/ml/gdb-patches/2008-05/msg00537.html
but has not yet been committed, prevent 
successful compilation of gdb configured with the two options
above :(

For cygwin native gdb, the testsuite results show no change to the previous
version.


                === gdb Summary ===

! # of expected passes          10791
! # of unexpected failures      581
  # of expected failures                59
  # of unknown successes                2
! # of known failures           22
  # of unresolved testcases     40
  # of untested testcases               14
  # of unsupported tests                23
! /usr/local/src/gdbcvs/build-bare/gdb/testsuite/../../gdb/gdb version
6.8.50.2
0080606-cvs -nx

--- 12103,12115 ----

                === gdb Summary ===

! # of expected passes          10954
! # of unexpected failures      386
  # of expected failures                59
  # of unknown successes                2
! # of known failures           23
  # of unresolved testcases     40
  # of untested testcases               14
  # of unsupported tests                23
! /usr/local/src/gdbcvs/build-bare/gdb/testsuite/../../gdb/gdb version
6.8.50.2
0080604-cvs -nx


ChangeLog entry:

2008-06-07  Pedro Alves  <pedro_alves@portugalmail.pt>
	Pierre Muller  <muller@ics.u-strasbg.fr>

	* gdbarch.sh (gdbarch_skip_main_prologue): New.
	* gdbarch.h, gdbarch.c: Regenerate.
	* i386-tdep.h (i386_skip_main_prologue): Declare.
	* i386-tdep.c (i386_skip_main_prologue): New.
	* i386-cygwin-tdep.c (i386_cygwin_init_abi): Register 
	i386_skip_main_prologue as gdbarch_skip_main_prologue gdbarch
callback.
	* symtab.c (find_function_start_sal): When pc points at the "main"
	function, call gdbarch_skip_main_prologue.


Index: gdb/gdbarch.c
===================================================================
RCS file: /cvs/src/src/gdb/gdbarch.c,v
retrieving revision 1.428
diff -u -p -r1.428 gdbarch.c
--- gdb/gdbarch.c	24 May 2008 16:32:01 -0000	1.428
+++ gdb/gdbarch.c	6 Jun 2008 10:54:57 -0000
@@ -183,6 +183,7 @@ struct gdbarch
   gdbarch_integer_to_address_ftype *integer_to_address;
   gdbarch_return_value_ftype *return_value;
   gdbarch_skip_prologue_ftype *skip_prologue;
+  gdbarch_skip_main_prologue_ftype *skip_main_prologue;
   gdbarch_inner_than_ftype *inner_than;
   gdbarch_breakpoint_from_pc_ftype *breakpoint_from_pc;
   gdbarch_adjust_breakpoint_address_ftype *adjust_breakpoint_address;
@@ -313,6 +314,7 @@ struct gdbarch startup_gdbarch =
   0,  /* integer_to_address */
   0,  /* return_value */
   0,  /* skip_prologue */
+  0,  /* skip_main_prologue */
   0,  /* inner_than */
   0,  /* breakpoint_from_pc */
   0,  /* adjust_breakpoint_address */
@@ -561,6 +563,7 @@ verify_gdbarch (struct gdbarch *gdbarch)
   /* Skip verify of return_value, has predicate */
   if (gdbarch->skip_prologue == 0)
     fprintf_unfiltered (log, "\n\tskip_prologue");
+  /* Skip verify of skip_main_prologue, has predicate */
   if (gdbarch->inner_than == 0)
     fprintf_unfiltered (log, "\n\tinner_than");
   if (gdbarch->breakpoint_from_pc == 0)
@@ -999,6 +1002,12 @@ gdbarch_dump (struct gdbarch *gdbarch, s
                       "gdbarch_dump: single_step_through_delay =
<0x%lx>\n",
                       (long) gdbarch->single_step_through_delay);
   fprintf_unfiltered (file,
+                      "gdbarch_dump: gdbarch_skip_main_prologue_p() =
%d\n",
+                      gdbarch_skip_main_prologue_p (gdbarch));
+  fprintf_unfiltered (file,
+                      "gdbarch_dump: skip_main_prologue = <0x%lx>\n",
+                      (long) gdbarch->skip_main_prologue);
+  fprintf_unfiltered (file,
                       "gdbarch_dump: gdbarch_skip_permanent_breakpoint_p()
= %d\n",
                       gdbarch_skip_permanent_breakpoint_p (gdbarch));
   fprintf_unfiltered (file,
@@ -2123,6 +2132,30 @@ set_gdbarch_skip_prologue (struct gdbarc
 }
 
 int
+gdbarch_skip_main_prologue_p (struct gdbarch *gdbarch)
+{
+  gdb_assert (gdbarch != NULL);
+  return gdbarch->skip_main_prologue != NULL;
+}
+
+CORE_ADDR
+gdbarch_skip_main_prologue (struct gdbarch *gdbarch, CORE_ADDR ip)
+{
+  gdb_assert (gdbarch != NULL);
+  gdb_assert (gdbarch->skip_main_prologue != NULL);
+  if (gdbarch_debug >= 2)
+    fprintf_unfiltered (gdb_stdlog, "gdbarch_skip_main_prologue called\n");
+  return gdbarch->skip_main_prologue (gdbarch, ip);
+}
+
+void
+set_gdbarch_skip_main_prologue (struct gdbarch *gdbarch,
+                              gdbarch_skip_main_prologue_ftype
skip_main_prologue)
+{
+  gdbarch->skip_main_prologue = skip_main_prologue;
+}
+
+int
 gdbarch_inner_than (struct gdbarch *gdbarch, CORE_ADDR lhs, CORE_ADDR rhs)
 {
   gdb_assert (gdbarch != NULL);
Index: gdb/gdbarch.h
===================================================================
RCS file: /cvs/src/src/gdb/gdbarch.h,v
retrieving revision 1.383
diff -u -p -r1.383 gdbarch.h
--- gdb/gdbarch.h	24 May 2008 16:32:01 -0000	1.383
+++ gdb/gdbarch.h	6 Jun 2008 10:54:57 -0000
@@ -379,6 +379,12 @@ typedef CORE_ADDR (gdbarch_skip_prologue
 extern CORE_ADDR gdbarch_skip_prologue (struct gdbarch *gdbarch, CORE_ADDR
ip);
 extern void set_gdbarch_skip_prologue (struct gdbarch *gdbarch,
gdbarch_skip_prologue_ftype *skip_prologue);
 
+extern int gdbarch_skip_main_prologue_p (struct gdbarch *gdbarch);
+
+typedef CORE_ADDR (gdbarch_skip_main_prologue_ftype) (struct gdbarch
*gdbarch, CORE_ADDR ip);
+extern CORE_ADDR gdbarch_skip_main_prologue (struct gdbarch *gdbarch,
CORE_ADDR ip);
+extern void set_gdbarch_skip_main_prologue (struct gdbarch *gdbarch,
gdbarch_skip_main_prologue_ftype *skip_main_prologue);
+
 typedef int (gdbarch_inner_than_ftype) (CORE_ADDR lhs, CORE_ADDR rhs);
 extern int gdbarch_inner_than (struct gdbarch *gdbarch, CORE_ADDR lhs,
CORE_ADDR rhs);
 extern void set_gdbarch_inner_than (struct gdbarch *gdbarch,
gdbarch_inner_than_ftype *inner_than);
Index: gdb/gdbarch.sh
===================================================================
RCS file: /cvs/src/src/gdb/gdbarch.sh,v
retrieving revision 1.469
diff -u -p -r1.469 gdbarch.sh
--- gdb/gdbarch.sh	24 May 2008 16:32:01 -0000	1.469
+++ gdb/gdbarch.sh	6 Jun 2008 10:54:58 -0000
@@ -482,6 +482,7 @@ M:CORE_ADDR:integer_to_address:struct ty
 M:enum return_value_convention:return_value:struct type *functype, struct
type *valtype, struct regcache *regcache, gdb_byte *readbuf, const gdb_byte
*writebuf:functype, valtype, regcache, readbuf, writebuf
 
 m:CORE_ADDR:skip_prologue:CORE_ADDR ip:ip:0:0
+M:CORE_ADDR:skip_main_prologue:CORE_ADDR ip:ip
 f:int:inner_than:CORE_ADDR lhs, CORE_ADDR rhs:lhs, rhs:0:0
 m:const gdb_byte *:breakpoint_from_pc:CORE_ADDR *pcptr, int *lenptr:pcptr,
lenptr::0:
 M:CORE_ADDR:adjust_breakpoint_address:CORE_ADDR bpaddr:bpaddr
Index: gdb/i386-cygwin-tdep.c
===================================================================
RCS file: /cvs/src/src/gdb/i386-cygwin-tdep.c,v
retrieving revision 1.16
diff -u -p -r1.16 i386-cygwin-tdep.c
--- gdb/i386-cygwin-tdep.c	1 Jan 2008 22:53:10 -0000	1.16
+++ gdb/i386-cygwin-tdep.c	6 Jun 2008 10:54:58 -0000
@@ -227,6 +227,8 @@ i386_cygwin_init_abi (struct gdbarch_inf
 
   set_gdbarch_skip_trampoline_code (gdbarch,
i386_cygwin_skip_trampoline_code);
 
+  set_gdbarch_skip_main_prologue (gdbarch, i386_skip_main_prologue);
+
   tdep->struct_return = reg_struct_return;
 
   tdep->gregset_reg_offset = i386_win32_gregset_reg_offset;
Index: gdb/i386-tdep.c
===================================================================
RCS file: /cvs/src/src/gdb/i386-tdep.c,v
retrieving revision 1.258
diff -u -p -r1.258 i386-tdep.c
--- gdb/i386-tdep.c	16 May 2008 00:27:23 -0000	1.258
+++ gdb/i386-tdep.c	6 Jun 2008 10:54:59 -0000
@@ -1160,6 +1160,38 @@ i386_skip_prologue (struct gdbarch *gdba
   return pc;
 }
 
+/* Check that the code pointed to by PC corresponds to a call to
+   __main, skip it if so.  Return PC otherwise.  */
+
+CORE_ADDR
+i386_skip_main_prologue (struct gdbarch *gdbarch, CORE_ADDR pc)
+{
+  gdb_byte op;
+
+  target_read_memory (pc, &op, 1);
+  if (op == 0xe8)
+    {
+      gdb_byte buf[4];
+
+      if (target_read_memory (pc + 1, buf, sizeof buf) == 0)
+ 	{
+	  /* Make sure address is computed correctly as a 32bit
+	     integer even if CORE_ADDR is 64 bit wide.  */
+ 	  struct minimal_symbol *s;
+ 	  CORE_ADDR call_dest = pc + 5 + extract_signed_integer (buf, 4);
+
+	  call_dest = call_dest & 0xffffffffU;
+ 	  s = lookup_minimal_symbol_by_pc (call_dest);
+ 	  if (s != NULL
+ 	      && SYMBOL_LINKAGE_NAME (s) != NULL
+ 	      && strcmp (SYMBOL_LINKAGE_NAME (s), "__main") == 0)
+ 	    pc += 5;
+ 	}
+    }
+
+  return pc;
+}
+
 /* This function is 64-bit safe.  */
 
 static CORE_ADDR
Index: gdb/i386-tdep.h
===================================================================
RCS file: /cvs/src/src/gdb/i386-tdep.h,v
retrieving revision 1.54
diff -u -p -r1.54 i386-tdep.h
--- gdb/i386-tdep.h	2 May 2008 16:49:54 -0000	1.54
+++ gdb/i386-tdep.h	6 Jun 2008 10:54:59 -0000
@@ -170,6 +170,7 @@ extern struct type *i386_sse_type (struc
 
 /* Functions exported from i386-tdep.c.  */
 extern CORE_ADDR i386_pe_skip_trampoline_code (CORE_ADDR pc, char *name);
+extern CORE_ADDR i386_skip_main_prologue (struct gdbarch *gdbarch,
CORE_ADDR pc);
 
 /* Return the name of register REGNUM.  */
 extern char const *i386_register_name (struct gdbarch * gdbarch, int
regnum);
Index: gdb/symtab.c
===================================================================
RCS file: /cvs/src/src/gdb/symtab.c,v
retrieving revision 1.189
diff -u -p -r1.189 symtab.c
--- gdb/symtab.c	27 May 2008 19:29:51 -0000	1.189
+++ gdb/symtab.c	6 Jun 2008 10:55:00 -0000
@@ -2617,6 +2617,21 @@ find_function_start_sal (struct symbol *
       /* Recalculate the line number (might not be N+1).  */
       sal = find_pc_sect_line (pc, SYMBOL_BFD_SECTION (sym), 0);
     }
+
+  /* On targets with executable formats that don't have a concept of
+     constructors (ELF with .init has, PE doesn't), gcc emits a call
+     to `__main' in `main' between the prologue and before user
+     code.  */
+  if (funfirstline
+      && gdbarch_skip_main_prologue_p (current_gdbarch)
+      && SYMBOL_LINKAGE_NAME (sym)
+      && strcmp (SYMBOL_LINKAGE_NAME (sym), "main") == 0)
+    {
+      pc = gdbarch_skip_main_prologue (current_gdbarch, pc);
+      /* Recalculate the line number (might not be N+1).  */
+      sal = find_pc_sect_line (pc, SYMBOL_BFD_SECTION (sym), 0);
+    }
+
   sal.pc = pc;
 
   return sal;

Re: [RFA] New skip __main version

[Mark Kettenis]

> From: "Pierre Muller" <muller@ics.u-strasbg.fr>
> Date: Sat, 7 Jun 2008 09:48:15 +0200
> 
> Here is a new version of this patch:

Looks ok to me.

> ChangeLog entry:
> 
> 2008-06-07  Pedro Alves  <pedro_alves@portugalmail.pt>
> 	Pierre Muller  <muller@ics.u-strasbg.fr>
> 
> 	* gdbarch.sh (gdbarch_skip_main_prologue): New.
> 	* gdbarch.h, gdbarch.c: Regenerate.
> 	* i386-tdep.h (i386_skip_main_prologue): Declare.
> 	* i386-tdep.c (i386_skip_main_prologue): New.
> 	* i386-cygwin-tdep.c (i386_cygwin_init_abi): Register 
> 	i386_skip_main_prologue as gdbarch_skip_main_prologue gdbarch
> callback.
> 	* symtab.c (find_function_start_sal): When pc points at the "main"
> 	function, call gdbarch_skip_main_prologue.

RE: [RFA] New skip __main version

[Pierre Muller]

I took the liberty to interpret this answer from Mark
as an approval and committed this patch.

Thanks to all,


Pierre Muller
Pascal language support maintainer for GDB



-----Message d'origine-----
De : gdb-patches-owner@sourceware.org
[mailto:gdb-patches-owner@sourceware.org] De la part de Mark Kettenis
Envoyé : Tuesday, June 10, 2008 3:49 PM
À : muller@ics.u-strasbg.fr
Cc : pedro@codesourcery.com; gdb-patches@sourceware.org; drow@false.org
Objet : Re: [RFA] New skip __main version

> From: "Pierre Muller" <muller@ics.u-strasbg.fr>
> Date: Sat, 7 Jun 2008 09:48:15 +0200
> 
> Here is a new version of this patch:

Looks ok to me.

> ChangeLog entry:
> 
> 2008-06-07  Pedro Alves  <pedro_alves@portugalmail.pt>
> 	Pierre Muller  <muller@ics.u-strasbg.fr>
> 
> 	* gdbarch.sh (gdbarch_skip_main_prologue): New.
> 	* gdbarch.h, gdbarch.c: Regenerate.
> 	* i386-tdep.h (i386_skip_main_prologue): Declare.
> 	* i386-tdep.c (i386_skip_main_prologue): New.
> 	* i386-cygwin-tdep.c (i386_cygwin_init_abi): Register 
> 	i386_skip_main_prologue as gdbarch_skip_main_prologue gdbarch
> callback.
> 	* symtab.c (find_function_start_sal): When pc points at the "main"
> 	function, call gdbarch_skip_main_prologue.