Re: [PATCH] Support DWARF 2/3 on i?86 PE

Re: [PATCH] Support DWARF 2/3 on i?86 PE

[Nick Clifton]

Hi Brian,

 Sorry for the long delay in approving this patch.  I have however now
 looked at it, approved it and applied it to the sources.

Cheers
        Nick

> bfd/ChangeLog:
> 2004-04-14  Brian Ford  <ford@vss.fsi.com>
> 
> 	* bfd.c (bfd_get_sign_extend_vma): Add pe[i]-i386 case to DJGPP hack.
> 	* coffcode.h (DOT_DEBUG, GNU_LINKONCE_WI): Define.
> 	[!COFF_WITH_PE] (sec_to_styp_flags, styp_to_sec_flags): Use them.
> 	(coff_compute_section_file_positions) [RS6000COFF_C]: Likewise.
> 	[COFF_WITH_PE] (sec_to_styp_flags): Handle DWARF 2/3 .debug* and
> 	.gnu.linkonce.wi. sections.
> 	* pe-i386.c (COFF_SUPPORT_GNU_LINKONCE): Define.
> 	(COFF_SECTION_ALIGNMENT_ENTRIES): Add entries for .debug and
> 	.gnu.linkonce.wi..
> 	* pei-i386.c (COFF_SUPPORT_GNU_LINKONCE): Likewise.
> 	(COFF_SECTION_ALIGNMENT_ENTRIES): Likewise.
> 
> gas/ChangeLog:
> 2004-04-14  Brian Ford  <ford@vss.fsi.com>
> 
> 	* dwarf2dbg.c (dwarf2_finish): Add SEC_DEBUGGING to section flags.
> 
> ld/ChangeLog:
> 2004-04-14  Brian Ford  <ford@vss.fsi.com>
> 
> 	* scripttempl/pe.sc: Handle .debug* and .gnu.linkonce.wi.* sections
> 	for DWARF 2/3.  Update stab section's syntax.

Approved.

Re: [PATCH] Support DWARF 2/3 on i?86 PE

[Alan Modra]

On Fri, Apr 16, 2004 at 10:20:18AM -0500, Brian Ford wrote:
> I assume you mean remove the leading * from the second line and end with
> two spaces before the */?  Sorry,  I know better.  But, I wanted to
> clarify and make sure I wasn't missing anything else.

Yes, that was all that was wrong with the formatting.

> > Hmm, actually the comment is wrong.  You want:
> >
> > 	/* The first non-zero section filepos is the header size.
> > 	   Sections without contents will have a filepos of 0.  */
> 
> I agree that says *approximately* the same thing, so I have changed it
> accordingly.  But, I still don't understand why you think the original
> comment was wrong?
> 
> "Orphaned and alloc only sections may have a filepos of 0."
> 
> Orphaned sections may have contents.  They just don't end up in the output
> file.  Also, alloc only implies no contents.

The comment as I wrote it states a stronger condition.  The part that
was wrong in the original comment was the implication that orphan
sections are special.  They aren't, apart from (by definition) not being
mentioned in the linker script.  Whether they have a filepos or not has
no particular relation to them being orphan sections.

> Ok, but as I said in a previous email, I don't have a sources.redhat.com
> account, or binutils write access.  Did you want to check these two in, or
> do you want to "sponsor" me for "Write After Approval" access?  I'll
> submit the appropriate form if so.

I'll check them in, and you can also apply for write after approval with
me as sponsor.

-- 
Alan Modra
IBM OzLabs - Linux Technology Centre

Re: [PATCH] Support DWARF 2/3 on i?86 PE

[Brian Ford]

On Fri, 16 Apr 2004, Alan Modra wrote:

> On Thu, Apr 15, 2004 at 10:35:54PM -0500, Brian Ford wrote:
> > http://sources.redhat.com/ml/binutils/2004-04/msg00169.html
>
> This patch is OK for mainline, provided you fix the comment formatting.

Thanks.

I assume you mean remove the leading * from the second line and end with
two spaces before the */?  Sorry,  I know better.  But, I wanted to
clarify and make sure I wasn't missing anything else.

> Hmm, actually the comment is wrong.  You want:
>
> 	/* The first non-zero section filepos is the header size.
> 	   Sections without contents will have a filepos of 0.  */

I agree that says *approximately* the same thing, so I have changed it
accordingly.  But, I still don't understand why you think the original
comment was wrong?

"Orphaned and alloc only sections may have a filepos of 0."

Orphaned sections may have contents.  They just don't end up in the output
file.  Also, alloc only implies no contents.

Sorry, I just don't get it.

> > http://sources.redhat.com/ml/binutils/2004-03/msg00542.html
>
> This is also OK.

Ok, but as I said in a previous email, I don't have a sources.redhat.com
account, or binutils write access.  Did you want to check these two in, or
do you want to "sponsor" me for "Write After Approval" access?  I'll
submit the appropriate form if so.

Thanks again.

-- 
Brian Ford
Senior Realtime Software Engineer
VITAL - Visual Simulation Systems
FlightSafety International
Phone: 314-551-8460
Fax:   314-551-8444

Re: [PATCH] Support DWARF 2/3 on i?86 PE

[Alan Modra]

On Thu, Apr 15, 2004 at 10:35:54PM -0500, Brian Ford wrote:
> http://sources.redhat.com/ml/binutils/2004-04/msg00169.html

This patch is OK for mainline, provided you fix the comment formatting.
Hmm, actually the comment is wrong.  You want:

	/* The first non-zero section filepos is the header size.
	   Sections without contents will have a filepos of 0.  */

> http://sources.redhat.com/ml/binutils/2004-03/msg00542.html

This is also OK.

-- 
Alan Modra
IBM OzLabs - Linux Technology Centre

Re: [PATCH] Support DWARF 2/3 on i?86 PE

[Brian Ford]

On Fri, 16 Apr 2004, Alan Modra wrote:

> On Thu, Apr 15, 2004 at 09:49:41PM -0500, Brian Ford wrote:
> > Just a friendly reminder :), the following three patches are still pending
> > and fix bugs uncovered by these patches.
> >
> > http://sources.redhat.com/ml/binutils/2004-04/msg00279.html
> > http://sources.redhat.com/ml/binutils/2004-04/msg00279.html
> >
> > Ok, this one isn't *that* related, but it's still pending ;-).
> >
> > http://sources.redhat.com/ml/binutils/2004-04/msg00279.html
>
> The same patch 3 times?  :)  I've committed it to mainline.

Ugh..., no.  Sorry.  Buggy Cygwin/X server cut-and-paste.

That should be:

http://sources.redhat.com/ml/binutils/2004-04/msg00169.html

followed by:

http://sources.redhat.com/ml/binutils/2004-03/msg00542.html

-- 
Brian Ford
Senior Realtime Software Engineer
VITAL - Visual Simulation Systems
FlightSafety International
Phone: 314-551-8460
Fax:   314-551-8444

[PATCH] Support DWARF 2/3 on i?86 PE

[Brian Ford]

[-- Attachment #1: Type: TEXT/PLAIN, Size: 818 bytes --]

I guess the subject says it all.

After this and the previous .secrel32 patch are applied, I'll submit the
gcc patch to enable DWARF 2 on PE (Cygwin/Mingw) using them.

Just a friendly reminder :), the following three patches are still pending
and fix bugs uncovered by these patches.

http://sources.redhat.com/ml/binutils/2004-04/msg00279.html
http://sources.redhat.com/ml/binutils/2004-04/msg00279.html

Ok, this one isn't *that* related, but it's still pending ;-).

http://sources.redhat.com/ml/binutils/2004-04/msg00279.html

I know Nick has been out for some time and is probably trying to dig out
slowly.  No rush :-).

Thanks in advance again for the review.

-- 
Brian Ford
Senior Realtime Software Engineer
VITAL - Visual Simulation Systems
FlightSafety International
Phone: 314-551-8460
Fax:   314-551-8444

[-- Attachment #2: Type: TEXT/PLAIN, Size: 973 bytes --]

bfd/ChangeLog:

2004-04-14  Brian Ford  <ford@vss.fsi.com>

	* bfd.c (bfd_get_sign_extend_vma): Add pe[i]-i386 case to DJGPP hack.
	* coffcode.h (DOT_DEBUG, GNU_LINKONCE_WI): Define.
	[!COFF_WITH_PE] (sec_to_styp_flags, styp_to_sec_flags): Use them.
	(coff_compute_section_file_positions) [RS6000COFF_C]: Likewise.
	[COFF_WITH_PE] (sec_to_styp_flags): Handle DWARF 2/3 .debug* and
	.gnu.linkonce.wi. sections.
	* pe-i386.c (COFF_SUPPORT_GNU_LINKONCE): Define.
	(COFF_SECTION_ALIGNMENT_ENTRIES): Add entries for .debug and
	.gnu.linkonce.wi..
	* pei-i386.c (COFF_SUPPORT_GNU_LINKONCE): Likewise.
	(COFF_SECTION_ALIGNMENT_ENTRIES): Likewise.

gas/ChangeLog:

2004-04-14  Brian Ford  <ford@vss.fsi.com>

	* dwarf2dbg.c (dwarf2_finish): Add SEC_DEBUGGING to section flags.

ld/ChangeLog:

2004-04-14  Brian Ford  <ford@vss.fsi.com>

	* scripttempl/pe.sc: Handle .debug* and .gnu.linkonce.wi.* sections
	for DWARF 2/3.  Update stab section's syntax.

[-- Attachment #3: Type: TEXT/PLAIN, Size: 10166 bytes --]

Index: bfd/bfd.c
===================================================================
RCS file: /cvs/src/src/bfd/bfd.c,v
retrieving revision 1.56
diff -u -p -r1.56 bfd.c
--- bfd/bfd.c	15 Apr 2004 08:48:56 -0000	1.56
+++ bfd/bfd.c	16 Apr 2004 01:44:30 -0000
@@ -769,12 +769,14 @@ bfd_get_sign_extend_vma (bfd *abfd)
 
   name = bfd_get_target (abfd);
 
-  /* Return a proper value for DJGPP COFF (an x86 COFF variant).
+  /* Return a proper value for DJGPP & PE COFF (x86 COFF variants).
      This function is required for DWARF2 support, but there is
      no place to store this information in the COFF back end.
      Should enough other COFF targets add support for DWARF2,
      a place will have to be found.  Until then, this hack will do.  */
-  if (strncmp (name, "coff-go32", sizeof ("coff-go32") - 1) == 0)
+  if (strncmp (name, "coff-go32", sizeof ("coff-go32") - 1) == 0
+      || strcmp (name, "pe-i386") == 0
+      || strcmp (name, "pei-i386") == 0)
     return 1;
 
   bfd_set_error (bfd_error_wrong_format);
Index: bfd/coffcode.h
===================================================================
RCS file: /cvs/src/src/bfd/coffcode.h,v
retrieving revision 1.101
diff -u -p -r1.101 coffcode.h
--- bfd/coffcode.h	19 Dec 2003 10:01:59 -0000	1.101
+++ bfd/coffcode.h	16 Apr 2004 01:44:32 -0000
@@ -309,6 +309,9 @@ CODE_FRAGMENT
 
 #define STRING_SIZE_SIZE (4)
 
+#define DOT_DEBUG	".debug"
+#define GNU_LINKONCE_WI ".gnu.linkonce.wi."
+
 static long sec_to_styp_flags
   PARAMS ((const char *, flagword));
 static bfd_boolean styp_to_sec_flags
@@ -428,7 +431,7 @@ sec_to_styp_flags (sec_name, sec_flags)
       styp_flags = STYP_LIT;
 #endif /* _LIT */
     }
-  else if (!strncmp (sec_name, ".debug", 6))
+  else if (!strncmp (sec_name, DOT_DEBUG, sizeof (DOT_DEBUG) - 1))
     {
       /* Handle the XCOFF debug section and DWARF2 debug sections.  */
       if (!sec_name[6])
@@ -441,7 +444,7 @@ sec_to_styp_flags (sec_name, sec_flags)
       styp_flags = STYP_DEBUG_INFO;
     }
 #ifdef COFF_LONG_SECTION_NAMES
-  else if (!strncmp (sec_name, ".gnu.linkonce.wi.", 17))
+  else if (!strncmp (sec_name, GNU_LINKONCE_WI, sizeof (GNU_LINKONCE_WI) - 1))
     {
       styp_flags = STYP_DEBUG_INFO;
     }
@@ -518,7 +521,7 @@ sec_to_styp_flags (sec_name, sec_flags)
 
 static long
 sec_to_styp_flags (sec_name, sec_flags)
-     const char *sec_name ATTRIBUTE_UNUSED;
+     const char *sec_name;
      flagword sec_flags;
 {
   long styp_flags = 0;
@@ -531,6 +534,11 @@ sec_to_styp_flags (sec_name, sec_flags)
      PE files.  The STYP_* flags and the IMAGE_SCN_* flags overlap,
      but there are more IMAGE_SCN_* flags.  */
 
+  /* FIXME: There is no gas syntax to specify the debug section flag.  */
+  if (strncmp(sec_name, DOT_DEBUG, sizeof (DOT_DEBUG) - 1) == 0
+      || strncmp(sec_name, GNU_LINKONCE_WI, sizeof (GNU_LINKONCE_WI) - 1) == 0)
+    sec_flags = SEC_READONLY | SEC_DEBUGGING;
+
   /* skip LOAD */
   /* READONLY later */
   /* skip RELOC */
@@ -675,12 +683,12 @@ styp_to_sec_flags (abfd, hdr, name, sect
 #endif
 	sec_flags |= SEC_ALLOC;
     }
-  else if (strncmp (name, ".debug", 6) == 0
+  else if (strncmp (name, DOT_DEBUG, sizeof (DOT_DEBUG) - 1) == 0
 #ifdef _COMMENT
 	   || strcmp (name, _COMMENT) == 0
 #endif
 #ifdef COFF_LONG_SECTION_NAMES
-	   || strncmp (name, ".gnu.linkonce.wi.", 17) == 0
+	   || strncmp (name, GNU_LINKONCE_WI, sizeof (GNU_LINKONCE_WI) - 1) == 0
 #endif
 	   || strncmp (name, ".stab", 5) == 0)
     {
@@ -3004,7 +3012,7 @@ coff_compute_section_file_positions (abf
 	{
 	  asection *dsec;
 
-	  dsec = bfd_make_section_old_way (abfd, ".debug");
+	  dsec = bfd_make_section_old_way (abfd, DOT_DEBUG);
 	  if (dsec == NULL)
 	    abort ();
 	  dsec->_raw_size = sz;
Index: bfd/pe-i386.c
===================================================================
RCS file: /cvs/src/src/bfd/pe-i386.c,v
retrieving revision 1.6
diff -u -p -r1.6 pe-i386.c
--- bfd/pe-i386.c	30 Nov 2002 08:39:40 -0000	1.6
+++ bfd/pe-i386.c	16 Apr 2004 01:44:33 -0000
@@ -26,6 +26,7 @@
 #define PCRELOFFSET TRUE
 #define TARGET_UNDERSCORE '_'
 #define COFF_LONG_SECTION_NAMES
+#define COFF_SUPPORT_GNU_LINKONCE
 #define COFF_LONG_FILENAMES
 
 #define COFF_SECTION_ALIGNMENT_ENTRIES \
@@ -38,6 +39,10 @@
 { COFF_SECTION_NAME_PARTIAL_MATCH (".idata"), \
   COFF_ALIGNMENT_FIELD_EMPTY, COFF_ALIGNMENT_FIELD_EMPTY, 2 }, \
 { COFF_SECTION_NAME_EXACT_MATCH (".pdata"), \
-  COFF_ALIGNMENT_FIELD_EMPTY, COFF_ALIGNMENT_FIELD_EMPTY, 2 }
+  COFF_ALIGNMENT_FIELD_EMPTY, COFF_ALIGNMENT_FIELD_EMPTY, 2 }, \
+{ COFF_SECTION_NAME_PARTIAL_MATCH (".debug"), \
+  COFF_ALIGNMENT_FIELD_EMPTY, COFF_ALIGNMENT_FIELD_EMPTY, 0 }, \
+{ COFF_SECTION_NAME_PARTIAL_MATCH (".gnu.linkonce.wi."), \
+  COFF_ALIGNMENT_FIELD_EMPTY, COFF_ALIGNMENT_FIELD_EMPTY, 0 }
 
 #include "coff-i386.c"
Index: bfd/pei-i386.c
===================================================================
RCS file: /cvs/src/src/bfd/pei-i386.c,v
retrieving revision 1.5
diff -u -p -r1.5 pei-i386.c
--- bfd/pei-i386.c	30 Nov 2002 08:39:40 -0000	1.5
+++ bfd/pei-i386.c	16 Apr 2004 01:44:33 -0000
@@ -27,6 +27,7 @@
 #define PCRELOFFSET TRUE
 #define TARGET_UNDERSCORE '_'
 #define COFF_LONG_SECTION_NAMES
+#define COFF_SUPPORT_GNU_LINKONCE
 #define COFF_LONG_FILENAMES
 
 #define COFF_SECTION_ALIGNMENT_ENTRIES \
@@ -39,6 +40,10 @@
 { COFF_SECTION_NAME_PARTIAL_MATCH (".idata"), \
   COFF_ALIGNMENT_FIELD_EMPTY, COFF_ALIGNMENT_FIELD_EMPTY, 2 }, \
 { COFF_SECTION_NAME_EXACT_MATCH (".pdata"), \
-  COFF_ALIGNMENT_FIELD_EMPTY, COFF_ALIGNMENT_FIELD_EMPTY, 2 }
+  COFF_ALIGNMENT_FIELD_EMPTY, COFF_ALIGNMENT_FIELD_EMPTY, 2 }, \
+{ COFF_SECTION_NAME_PARTIAL_MATCH (".debug"), \
+  COFF_ALIGNMENT_FIELD_EMPTY, COFF_ALIGNMENT_FIELD_EMPTY, 0 }, \
+{ COFF_SECTION_NAME_PARTIAL_MATCH (".gnu.linkonce.wi."), \
+  COFF_ALIGNMENT_FIELD_EMPTY, COFF_ALIGNMENT_FIELD_EMPTY, 0 }
 
 #include "coff-i386.c"
Index: gas/dwarf2dbg.c
===================================================================
RCS file: /cvs/src/src/gas/dwarf2dbg.c,v
retrieving revision 1.69
diff -u -p -r1.69 dwarf2dbg.c
--- gas/dwarf2dbg.c	13 Feb 2004 16:12:46 -0000	1.69
+++ gas/dwarf2dbg.c	16 Apr 2004 01:44:35 -0000
@@ -1371,7 +1371,7 @@ dwarf2_finish (void)
 
   /* Create and switch to the line number section.  */
   line_seg = subseg_new (".debug_line", 0);
-  bfd_set_section_flags (stdoutput, line_seg, SEC_READONLY);
+  bfd_set_section_flags (stdoutput, line_seg, SEC_READONLY | SEC_DEBUGGING);
 
   /* For each subsection, chain the debug entries together.  */
   for (s = all_segs; s; s = s->next)
@@ -1400,9 +1400,12 @@ dwarf2_finish (void)
       abbrev_seg = subseg_new (".debug_abbrev", 0);
       aranges_seg = subseg_new (".debug_aranges", 0);
 
-      bfd_set_section_flags (stdoutput, info_seg, SEC_READONLY);
-      bfd_set_section_flags (stdoutput, abbrev_seg, SEC_READONLY);
-      bfd_set_section_flags (stdoutput, aranges_seg, SEC_READONLY);
+      bfd_set_section_flags (stdoutput, info_seg,
+			     SEC_READONLY | SEC_DEBUGGING);
+      bfd_set_section_flags (stdoutput, abbrev_seg,
+			     SEC_READONLY | SEC_DEBUGGING);
+      bfd_set_section_flags (stdoutput, aranges_seg,
+			     SEC_READONLY | SEC_DEBUGGING);
 
       record_alignment (aranges_seg, ffs (2 * sizeof_address) - 1);
 
Index: ld/scripttempl/pe.sc
===================================================================
RCS file: /cvs/src/src/ld/scripttempl/pe.sc,v
retrieving revision 1.9
diff -u -p -r1.9 pe.sc
--- ld/scripttempl/pe.sc	10 Nov 2003 17:04:55 -0000	1.9
+++ ld/scripttempl/pe.sc	16 Apr 2004 01:44:44 -0000
@@ -159,13 +159,92 @@ SECTIONS
 
   .stab ${RELOCATING+BLOCK(__section_alignment__)} ${RELOCATING+(NOLOAD)} :
   {
-    [ .stab ]
+    *(.stab)
   }
 
   .stabstr ${RELOCATING+BLOCK(__section_alignment__)} ${RELOCATING+(NOLOAD)} :
   {
-    [ .stabstr ]
+    *(.stabstr)
   }
 
+  /* DWARF debug sections.
+     Symbols in the DWARF debugging sections are relative to the beginning
+     of the section.  Unlike other targets that fake this by putting the
+     section VMA at 0, the PE format will not allow it.
+  */
+     
+  /* DWARF 1.1 and DWARF 2 */
+  .debug_aranges ${RELOCATING+BLOCK(__section_alignment__)} ${RELOCATING+(NOLOAD)} :
+  {
+    *(.debug_aranges)
+  }
+
+  .debug_pubnames ${RELOCATING+BLOCK(__section_alignment__)} ${RELOCATING+(NOLOAD)} :
+  {
+    *(.debug_pubnames)
+  }
+
+  /* DWARF 2 */
+  .debug_info ${RELOCATING+BLOCK(__section_alignment__)} ${RELOCATING+(NOLOAD)} :
+  {
+    *(.debug_info) *(.gnu.linkonce.wi.*)
+  }
+
+  .debug_abbrev ${RELOCATING+BLOCK(__section_alignment__)} ${RELOCATING+(NOLOAD)} :
+  {
+    *(.debug_abbrev)
+  }
+
+  .debug_line ${RELOCATING+BLOCK(__section_alignment__)} ${RELOCATING+(NOLOAD)} :
+  {
+    *(.debug_line)
+  }
+
+  .debug_frame ${RELOCATING+BLOCK(__section_alignment__)} ${RELOCATING+(NOLOAD)} :
+  {
+    *(.debug_frame)
+  }
+
+  .debug_str ${RELOCATING+BLOCK(__section_alignment__)} ${RELOCATING+(NOLOAD)} :
+  {
+    *(.debug_str)
+  }
+
+  .debug_loc ${RELOCATING+BLOCK(__section_alignment__)} ${RELOCATING+(NOLOAD)} :
+  {
+    *(.debug_loc)
+  }
+
+  .debug_macinfo ${RELOCATING+BLOCK(__section_alignment__)} ${RELOCATING+(NOLOAD)} :
+  {
+    *(.debug_macinfo)
+  }
+
+  /* SGI/MIPS DWARF 2 extensions.  */
+  .debug_weaknames ${RELOCATING+BLOCK(__section_alignment__)} ${RELOCATING+(NOLOAD)} :
+  {
+    *(.debug_weaknames)
+  }
+
+  .debug_funcnames ${RELOCATING+BLOCK(__section_alignment__)} ${RELOCATING+(NOLOAD)} :
+  {
+    *(.debug_funcnames)
+  }
+
+  .debug_typenames ${RELOCATING+BLOCK(__section_alignment__)} ${RELOCATING+(NOLOAD)} :
+  {
+    *(.debug_typenames)
+  }
+
+  .debug_varnames ${RELOCATING+BLOCK(__section_alignment__)} ${RELOCATING+(NOLOAD)} :
+  {
+    *(.debug_varnames)
+  }
+
+  /* DWARF 3 */
+  .debug_ranges ${RELOCATING+BLOCK(__section_alignment__)} ${RELOCATING+(NOLOAD)} :
+  {
+    *(.debug_ranges)
+  }
 }
 EOF