llvm dev - Feb 2011 - [LLVMdev] [patch] Dwarf Debug info support for COFF object files

On Feb 24, 2011, at 11:36 AM, Devang Patel wrote:
> 
> On Feb 12, 2011, at 2:07 AM, Nathan Jeffords wrote:
> 
>> Hello All,
>> 
>> I have created a set of patches that get dwarf debugging support
working for the COFF object file. I also believe I have fixed what appears to be
a bug in how line info sections are referred to from the DW_TAG_compile_unit
DIE. I have run some basic tests, analyzed dumps of both the objects files and
the final executables, and run a test program against mingw-gdb and everything
looks to be in order.
>> 
>> Here is a short description of what the patches accomplish
>> 
>> die.patch:
>>  adds a new DIEValue type to represent a section relative label. (their
was already a type id specified so I provided a class modeled after DIELabel)
>> 
>> secrel-fixup.patch:
>>  creats a new target specific fixup type (reloc_coff_secrel32) to
represent COFF Section Relative relocations and updats the COFF object writer to
write it as COFF_IMAGE_REL_AMD64_SREL32
>> 
>> secrel-streamer.patch
>>  adds a new directive in the MCStreamer interface to allow the
AsmPrinter to emit a section relative label and provides implementations for all
existings Streamer (all but WinCOFF either forward or error on it)
>> 
>> secrel-dwarf.patch
>>  updates dwarf printing code to make use of the new directive & DIE
value where appropriate (this is where the bug fix is)
>> 
>> coff-debug.patch
>>  turns the dwarf output on in the X86/COFF AsmInfo classes
> 
> This patches look fine. I am curious, how are you testing debug info
support for COFF ?
Well, this patch fails following FE tests on darwin.
    LLVM :: FrontendC++/2006-11-06-StackTrace.cpp
    LLVM :: FrontendC++/2006-11-30-Pubnames.cpp
    LLVM :: FrontendC++/2010-08-31-ByValArg.cpp
    LLVM :: FrontendC/2009-02-17-BitField-dbg.c
    LLVM :: FrontendC/2010-01-14-StaticVariable.c
    LLVM :: FrontendC/2010-02-16-DbgVarScope.c
    LLVM :: FrontendObjC/2009-08-17-DebugInfo.m

-
Devang

On Feb 24, 2011, at 11:52 AM, Devang Patel wrote:
> 
> On Feb 24, 2011, at 11:36 AM, Devang Patel wrote:
> 
>> 
>> On Feb 12, 2011, at 2:07 AM, Nathan Jeffords wrote:
>> 
>>> Hello All,
>>> 
>>> I have created a set of patches that get dwarf debugging support
working for the COFF object file. I also believe I have fixed what appears to be
a bug in how line info sections are referred to from the DW_TAG_compile_unit
DIE. I have run some basic tests, analyzed dumps of both the objects files and
the final executables, and run a test program against mingw-gdb and everything
looks to be in order.
>>> 
>>> Here is a short description of what the patches accomplish
>>> 
>>> die.patch:
>>> adds a new DIEValue type to represent a section relative label.
(their was already a type id specified so I provided a class modeled after
DIELabel)
>>> 
>>> secrel-fixup.patch:
>>> creats a new target specific fixup type (reloc_coff_secrel32) to
represent COFF Section Relative relocations and updats the COFF object writer to
write it as COFF_IMAGE_REL_AMD64_SREL32
>>> 
>>> secrel-streamer.patch
>>> adds a new directive in the MCStreamer interface to allow the
AsmPrinter to emit a section relative label and provides implementations for all
existings Streamer (all but WinCOFF either forward or error on it)
>>> 
>>> secrel-dwarf.patch
>>> updates dwarf printing code to make use of the new directive &
DIE value where appropriate (this is where the bug fix is)
>>> 
>>> coff-debug.patch
>>> turns the dwarf output on in the X86/COFF AsmInfo classes
>> 
>> This patches look fine. I am curious, how are you testing debug info
support for COFF ?
> 
> Well, this patch fails following FE tests on darwin.
>    LLVM :: FrontendC++/2006-11-06-StackTrace.cpp
>    LLVM :: FrontendC++/2006-11-30-Pubnames.cpp
>    LLVM :: FrontendC++/2010-08-31-ByValArg.cpp
>    LLVM :: FrontendC/2009-02-17-BitField-dbg.c
>    LLVM :: FrontendC/2010-01-14-StaticVariable.c
>    LLVM :: FrontendC/2010-02-16-DbgVarScope.c
>    LLVM :: FrontendObjC/2009-08-17-DebugInfo.m
This is because of 

@@ -1903,7 +1912,8 @@
     addLabel(Die, dwarf::DW_AT_stmt_list, dwarf::DW_FORM_addr,
              Asm->GetTempSymbol("section_line"));
   else
-    addUInt(Die, dwarf::DW_AT_stmt_list, dwarf::DW_FORM_data4, 0);
+    addSectionOffset(Die, dwarf::DW_AT_stmt_list, dwarf::DW_FORM_addr,
+             Asm->GetTempSymbol("section_line"));
 
   if (!Dir.empty())
     addString(Die, dwarf::DW_AT_comp_dir, dwarf::DW_FORM_string, Dir);


You probably wanted  to do...

@@ -1904,8 +1913,8 @@
   // DW_AT_stmt_list is a offset of line number information for this
   // compile unit in debug_line section.
   if (Asm->MAI->doesDwarfUsesAbsoluteLabelForStmtList())
-    addLabel(Die, dwarf::DW_AT_stmt_list, dwarf::DW_FORM_addr,
-             Asm->GetTempSymbol("section_line"));
+    addSectionOffset(Die, dwarf::DW_AT_stmt_list, dwarf::DW_FORM_addr,
+                     Asm->GetTempSymbol("section_line"));
   else
     addUInt(Die, dwarf::DW_AT_stmt_list, dwarf::DW_FORM_data4, 0);
 

I updated this part and applied your patch. r126425
Thanks for the contribution!
-
Devang
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20110224/f64e8829/attachment.html>

Thanks for taking the time to review this and get it into the mainline.

[snip]

This is because of
>
> @@ -1903,7 +1912,8 @@
>      addLabel(Die, dwarf::DW_AT_stmt_list, dwarf::DW_FORM_addr,
>               Asm->GetTempSymbol("section_line"));
>    else
> -    addUInt(Die, dwarf::DW_AT_stmt_list, dwarf::DW_FORM_data4, 0);
> +    addSectionOffset(Die, dwarf::DW_AT_stmt_list, dwarf::DW_FORM_addr,
> +             Asm->GetTempSymbol("section_line"));
>
>    if (!Dir.empty())
>      addString(Die, dwarf::DW_AT_comp_dir, dwarf::DW_FORM_string, Dir);
>
>
> You probably wanted  to do...
>
> @@ -1904,8 +1913,8 @@
>    // DW_AT_stmt_list is a offset of line number information for this
>    // compile unit in debug_line section.
>    if (Asm->MAI->doesDwarfUsesAbsoluteLabelForStmtList())
> -    addLabel(Die, dwarf::DW_AT_stmt_list, dwarf::DW_FORM_addr,
> -             Asm->GetTempSymbol("section_line"));
> +    addSectionOffset(Die, dwarf::DW_AT_stmt_list, dwarf::DW_FORM_addr,
> +                     Asm->GetTempSymbol("section_line"));
>    else
>      addUInt(Die, dwarf::DW_AT_stmt_list, dwarf::DW_FORM_data4, 0);
>
>
I am not sure about this, Looking at it now I don't understand what the
return value of doesDwarfUsesAbsoluteLabelForStmtList is supposed to mean. I
know that when I allowed the COFF's MCAsmInfo to use this line:

     addUInt(Die, dwarf::DW_AT_stmt_list, dwarf::DW_FORM_data4, 0);

I got a relocation that caused an absolute pointer to the debug_line section
instead of the offset from the beginning of the section. This looked like a
bug too me. To me it doesn't matter: I can
change DwarfUsesAbsoluteLabelForStmtList to true for MCAsmInfoCOFF and get
the correct behavior for COFF output.

[snip]

-Nathan
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20110224/1b91218c/attachment.html>