[Bug 23137] New: dbghelp doesn't support DW_LNE_set_discriminator (gcc 4.6)

http://bugs.winehq.org/show_bug.cgi?id=23137

Summary: dbghelp doesn't support DW_LNE_set_discriminator (gcc
                    4.6)
           Product: Wine
           Version: 1.2-rc3
          Platform: x86-64
        OS/Version: Linux
            Status: NEW
          Keywords: download, source
          Severity: enhancement
          Priority: P2
         Component: dbghelp
        AssignedTo: [e-mail]
        ReportedBy: [e-mail]

I've noticed lately that my backtraces get filled with:
fixme:dbghelp_dwarf:dwarf2_parse_line_numbers Unsupported extended opcode 4
fixme:dbghelp_dwarf:dwarf2_parse_line_numbers Unsupported extended opcode 4
fixme:dbghelp_dwarf:dwarf2_parse_line_numbers Unsupported extended opcode 4
fixme:dbghelp_dwarf:dwarf2_parse_line_numbers Unsupported extended opcode 4
fixme:dbghelp_dwarf:dwarf2_parse_line_numbers Unsupported extended opcode 4
fixme:dbghelp_dwarf:dwarf2_parse_line_numbers Unsupported extended opcode 4
fixme:dbghelp_dwarf:dwarf2_parse_line_numbers Unsupported extended opcode 4
fixme:dbghelp_dwarf:dwarf2_parse_line_numbers Unsupported extended opcode 4

looking around a bit more, seems that is DW_LNE_set_discriminator. I looked
into it a bit more, and tried:
diff --git a/dlls/dbghelp/dwarf.c b/dlls/dbghelp/dwarf.c
index 6daf7d7..89a9dfb 100644
--- a/dlls/dbghelp/dwarf.c
+++ b/dlls/dbghelp/dwarf.c
@@ -2075,6 +2075,9 @@ static BOOL dwarf2_parse_line_numbers(const
dwarf2_section_t* sections,
                         dwarf2_leb128_as_unsigned(&traverse);
                         dwarf2_leb128_as_unsigned(&traverse);
                         break;
+                    case DW_LNE_set_discriminator:
+                        FIXME("not yet supported...\n");
+                        break;
                     default:
                         FIXME("Unsupported extended opcode %x\n", extopcode);
                         break;
diff --git a/dlls/dbghelp/dwarf.h b/dlls/dbghelp/dwarf.h
index 4ecb816..1ea17c0 100644
--- a/dlls/dbghelp/dwarf.h
+++ b/dlls/dbghelp/dwarf.h
@@ -421,6 +421,7 @@ enum dwarf_calling_convention
 #define DW_LNE_end_sequence             0x01
 #define DW_LNE_set_address              0x02
 #define DW_LNE_define_file              0x03
+#define DW_LNE_set_discriminator        0x04

#define DW_CIE_ID                       ~(0x0)

but that obviously doesn't do much, and makes the backtraces worse than
before...

[Bug 23137] dbghelp doesn't support DW_LNE_set_discriminator (gcc 4.6)

http://bugs.winehq.org/show_bug.cgi?id=23137

Eric Pouech  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]

--- Comment #1 from Eric Pouech   2010-06-12 04:25:50 ---
you must add a call to dwarf2_leb128_as_unsigned(&traverse);
otherwise you'll be out of sync with the opcode information
not sure though we'll to support this opcode, mainly to parse it correctly

A+

[Bug 23137] dbghelp doesn't support DW_LNE_set_discriminator (gcc 4.6)

http://bugs.winehq.org/show_bug.cgi?id=23137

--- Comment #2 from Eric Pouech   2010-06-25 15:12:12 ---
Created an attachment (id=29137)
 --> (http://bugs.winehq.org/attachment.cgi?id=29137)
Full patch

Does the attached (with the extension I suggested) work better ?

[Bug 23137] dbghelp doesn't support DW_LNE_set_discriminator (gcc 4.6)

http://bugs.winehq.org/show_bug.cgi?id=23137

--- Comment #3 from Austin English   2010-06-27 02:08:48 ---
(In reply to comment #2)
> Created an attachment (id=29137)

--> (http://bugs.winehq.org/attachment.cgi?id=29137) [details]
> Full patch
>
> Does the attached (with the extension I suggested) work better ?

I tried it before myself...all it does is change the fixme, there's still
hundreds of them spewed on every backtrace.

[Bug 23137] dbghelp doesn't support DW_LNE_set_discriminator (gcc 4.6)

http://bugs.winehq.org/show_bug.cgi?id=23137

--- Comment #4 from Eric Pouech   2010-06-27 02:40:25 ---
if you change the FIXME into a WARN, does the debugger behave correctly ?
or do you get FIXMEs from another sort?

[Bug 23137] dbghelp doesn't support DW_LNE_set_discriminator (gcc 4.6)

http://bugs.winehq.org/show_bug.cgi?id=23137

--- Comment #5 from Austin English   2010-06-27 13:27:45 ---
Created an attachment (id=29192)
 --> (http://bugs.winehq.org/attachment.cgi?id=29192)
backtrace example

It's better, but there's still quite a few fixme's.

[Bug 23137] dbghelp doesn't support DW_LNE_set_discriminator (gcc 4.6)

http://bugs.winehq.org/show_bug.cgi?id=23137

--- Comment #6 from Eric Pouech   2010-06-27 15:57:02 ---
the latest fixmes seem to come from two distinct issues:
- GCC seems now to be able to alias a local var to a global var ; dbghelp spits
out a warning in that case (that was needed to debug the stabs part). this will
have no nocive impact (except the console garbage)
- missing save/restore of context in CFA management. AJ did it in
ntdll/signal_x64_86.c ; it should be ported to dlls/dbghelp/dwarf.c

TODOs after 1.2

[Bug 23137] dbghelp doesn't support DW_LNE_set_discriminator (gcc 4.6)

http://bugs.winehq.org/show_bug.cgi?id=23137

--- Comment #7 from Eric Pouech   2010-07-05 14:39:51 ---
support for save/restore has been committed *before* 1.2
please retest and if appropriate close the bug

[Bug 23137] dbghelp doesn't support DW_LNE_set_discriminator (gcc 4.6)

http://bugs.winehq.org/show_bug.cgi?id=23137

--- Comment #8 from Austin English   2010-07-07 23:08:00 ---
(In reply to comment #7)
> support for save/restore has been committed *before* 1.2
> please retest and if appropriate close the bug

Looking great! These lines are still there:
fixme:dbghelp:elf_new_wine_thunks Duplicate in L"kernel32<elf>":
__wine_dbch_time<7ef001c0-00000010> __dbch<7ef001c0-0>
fixme:dbghelp:elf_new_wine_thunks Duplicate in L"kernel32<elf>":
__wine_dbch_ver<7ef001e0-00000010> __dbch<7ef001e0-0>

but otherwise, much more manageable. Do you want a separate bug report for that
issue?

[Bug 23137] dbghelp doesn't support DW_LNE_set_discriminator (gcc 4.6)

http://bugs.winehq.org/show_bug.cgi?id=23137

--- Comment #9 from Eric Pouech  2010-09-11 01:19:47 CDT ---
yes a separate bug report is better (it's another issue, that was hidden behind
this bug report)

[Bug 23137] dbghelp doesn't support DW_LNE_set_discriminator (gcc 4.6)

http://bugs.winehq.org/show_bug.cgi?id=23137

Austin English  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Blocks|                            |24359

[Bug 23137] dbghelp doesn't support DW_LNE_set_discriminator (gcc 4.6)

http://bugs.winehq.org/show_bug.cgi?id=23137

Austin English  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED

--- Comment #10 from Austin English  2010-09-11 15:49:11 CDT ---
Fixed. Second part is bug 24359.

[Bug 23137] dbghelp doesn't support DW_LNE_set_discriminator (gcc 4.6)

http://bugs.winehq.org/show_bug.cgi?id=23137

Alexandre Julliard  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |CLOSED

--- Comment #11 from Alexandre Julliard  2010-09-18 13:05:24 CDT ---
Closing bugs fixed in 1.3.3.