Thread overview
Assembler problem while trying to integrate D iompiler in GCC suite
Jun 07, 2005
Gerrit P. Haase
Jun 07, 2005
Gerrit P. Haase
Jun 07, 2005
Gerrit P. Haase
Jun 07, 2005
David Friedman
Jun 07, 2005
Gerrit P. Haase
Jun 07, 2005
Gerrit P. Haase
Jun 07, 2005
Gerrit P. Haase
Jun 07, 2005
Gerrit P. Haase
June 07, 2005
Hi all,

I'm trying to integrate the D frontend for GCC in the GCC 3.4.4 release,
I'm in contact with the maintainer David Friedman, he successfully
compiled the current (0.12.1) sources with MinGW and I expect that he
also has tested it with other supported platforms, however I get this
error now and I have not really an idea what the reason may be.  I found
one patch for gas for another platform:
http://sources.redhat.com/bugzilla/show_bug.cgi?id=857
http://sources.redhat.com/bugzilla/show_bug.cgi?id=936
http://sourceware.org/ml/binutils/2005-05/msg00289.html

However I don't know if this is related or if there may be another problem with the e.g. source.

What is new in the source regarding this problem is this patch which gives me the error cited below:

diff -cr d-0.12/README d-0.12.1/README
*** d-0.12/README	Sat May 28 18:00:56 2005
--- d-0.12.1/README	Sun Jun  5 19:51:22 2005
***************
*** 1,6 ****
! D Front End for GCC - Release 0.12

! Last update: May 24, 2005
  Supported Systems

      * GCC 3.3.x, 3.4.x
--- 1,6 ----
! D Front End for GCC - Release 0.12.1

! Last update: June 5, 2005
  Supported Systems

      * GCC 3.3.x, 3.4.x
***************
*** 17,25 ****

  Downloads

!     * Main package (Release 0.12)
!       http://home.earthlink.net/~dvdfrdmn/d/gdc-0.12.tgz
!       http://home.earthlink.net/~dvdfrdmn/d/gdc-0.12.tar.bz2
      * Build Instructions
        http://home.earthlink.net/~dvdfrdmn/d/INSTALL.html
        (or see INSTALL included in the main package)
--- 17,25 ----

  Downloads

!     * Main package (Release 0.12.1)
!       http://home.earthlink.net/~dvdfrdmn/d/gdc-0.12.1.tgz
!       http://home.earthlink.net/~dvdfrdmn/d/gdc-0.12.1.tar.bz2
      * Build Instructions
        http://home.earthlink.net/~dvdfrdmn/d/INSTALL.html
        (or see INSTALL included in the main package)
***************
*** 172,177 ****
--- 172,186 ----

  Changes

+ 0.12.1:
+
+     *  Fixes
+           o Compilation errors and crashes with -fpic
+           o Crashes with interfaces on Cygwin/MinGW
+           o Support the -mno-cygwin option
+
+ 0.12:
+
      *  Fixes
            o Various problems building MinGW in MSYS
            o Mango "endless output"
diff -cr d-0.12/d-codegen.cc d-0.12.1/d-codegen.cc
*** d-0.12/d-codegen.cc	Sat May 28 18:00:56 2005
--- d-0.12.1/d-codegen.cc	Sun Jun  5 19:27:36 2005
***************
*** 1743,1748 ****
--- 1743,1805 ----
      return call_exp;
  }

+
+ #if D_GCC_VER >= 34
+ #ifdef ASM_OUTPUT_DEF
+ static int thunk_labelno;
+
+ /* Create a static alias to function.  */
+
+ static tree
+ make_alias_for_thunk (tree function)
+ {
+   tree alias;
+   char buf[256];
+
+ #if defined (TARGET_IS_PE_COFF)
+   if (DECL_ONE_ONLY (function))
+     return function;
+ #endif
+
+   ASM_GENERATE_INTERNAL_LABEL (buf, "LTHUNK", thunk_labelno);
+   thunk_labelno++;
+   alias = build_decl (FUNCTION_DECL, get_identifier (buf),
+ 		      TREE_TYPE (function));
+   //DECL_LANG_SPECIFIC (alias) = DECL_LANG_SPECIFIC (function);
+   //cxx_dup_lang_specific_decl (alias);
+   DECL_CONTEXT (alias) = NULL;
+   TREE_READONLY (alias) = TREE_READONLY (function);
+   TREE_THIS_VOLATILE (alias) = TREE_THIS_VOLATILE (function);
+   TREE_PUBLIC (alias) = 0;
+   //DECL_INTERFACE_KNOWN (alias) = 1;
+   //DECL_NOT_REALLY_EXTERN (alias) = 1;
+   //DECL_THIS_STATIC (alias) = 1;
+   //DECL_SAVED_FUNCTION_DATA (alias) = NULL;
+   //DECL_DESTRUCTOR_P (alias) = 0;
+   //DECL_CONSTRUCTOR_P (alias) = 0;
+   //DECL_CLONED_FUNCTION (alias) = NULL_TREE;
+   DECL_EXTERNAL (alias) = 0;
+   DECL_ARTIFICIAL (alias) = 1;
+   DECL_NO_STATIC_CHAIN (alias) = 1;
+   //DECL_PENDING_INLINE_P (alias) = 0;
+   DECL_INLINE (alias) = 0;
+   DECL_DECLARED_INLINE_P (alias) = 0;
+   //DECL_DEFERRED_FN (alias) = 0;
+   //DECL_USE_TEMPLATE (alias) = 0;
+   //DECL_TEMPLATE_INSTANTIATED (alias) = 0;
+   //DECL_TEMPLATE_INFO (alias) = NULL;
+   //DECL_INITIAL (alias) = error_mark_node;
+   TREE_ADDRESSABLE (alias) = 1;
+   TREE_USED (alias) = 1;
+   SET_DECL_ASSEMBLER_NAME (alias, DECL_NAME (alias));
+   TREE_SYMBOL_REFERENCED (DECL_ASSEMBLER_NAME (alias)) = 1;
+   if (!flag_syntax_only)
+     assemble_alias (alias, DECL_ASSEMBLER_NAME (function));
+   return alias;
+ }
+ #endif
+ #endif
+
  // Can't output thunks while a function is being compiled. Called from
  // d-decls.cc (no cur_irs)
  static Array deferredThunks;
***************
*** 1787,1799 ****
  IRState::outputThunk(tree thunk_decl, tree target_decl, int offset)
  {
      int delta = -offset;
!     tree alias = target_decl;

      TREE_ADDRESSABLE(target_decl) = 1;
      TREE_USED(target_decl) = 1;
  #if D_GCC_VER >= 34
      DECL_VISIBILITY (thunk_decl) = DECL_VISIBILITY (target_decl);
! #ifdef ASM_OUTPUT_DEF
    if (targetm.have_named_sections)
      {
        resolve_unique_section (target_decl, 0, flag_function_sections);
--- 1844,1866 ----
  IRState::outputThunk(tree thunk_decl, tree target_decl, int offset)
  {
      int delta = -offset;
!     tree alias;
! #if D_GCC_VER >= 34
! #ifdef ASM_OUTPUT_DEF
!     alias = make_alias_for_thunk(target_decl);
! #else
!     alias = target_decl;
! #endif
! #else
!     alias = target_decl;
! #endif

      TREE_ADDRESSABLE(target_decl) = 1;
      TREE_USED(target_decl) = 1;
  #if D_GCC_VER >= 34
      DECL_VISIBILITY (thunk_decl) = DECL_VISIBILITY (target_decl);
! #if defined (ASM_OUTPUT_DEF) \
!   && !defined (TARGET_IS_PE_COFF)
    if (targetm.have_named_sections)
      {
        resolve_unique_section (target_decl, 0, flag_function_sections);
diff -cr d-0.12/d-glue.cc d-0.12.1/d-glue.cc
*** d-0.12/d-glue.cc	Thu May 26 21:05:04 2005
--- d-0.12.1/d-glue.cc	Sun Jun  5 19:27:36 2005
***************
*** 2361,2366 ****
--- 2361,2367 ----
  		build_pointer_type( mod_ref_type ));
  	    dkeep(the_mod_ref);
  	    DECL_EXTERNAL( the_mod_ref ) = 1;
+ 	    TREE_PUBLIC( the_mod_ref ) = 1;

  	    tree m1 = build(MODIFY_EXPR, void_type_node,
  		build(COMPONENT_REF, TREE_TYPE(f0), our_mod_ref, f0),



Gives me this error now:

$ /gcc/gcc-3.4.4/gcc-3.4.4-1/.build/gcc/gdc -save-temps -B/gcc/gcc-3.4.4/gcc-3.4.4-1/.build/gcc/ -B/usr/i686-pc-cygwin/bin/ -B/usr/i686-pc-cy>
gdc: warning: -pipe ignored because -save-temps specified
socketstream.s: Assembler messages:
socketstream.s:1416: Error: Local symbol `LTHUNK0' can't be equated to undefined symbol `__D3std6stream6Stream9readExactFPvkZv'
socketstream.s:1416: Error: Local symbol `LTHUNK1' can't be equated to undefined symbol `__D3std6stream6Stream4readFAhZk'
socketstream.s:1416: Error: Local symbol `LTHUNK2' can't be equated to undefined symbol `__D3std6stream6Stream4readFJgZv'
socketstream.s:1416: Error: Local symbol `LTHUNK3' can't be equated to undefined symbol `__D3std6stream6Stream4readFJhZv'
socketstream.s:1416: Error: Local symbol `LTHUNK4' can't be equated to undefined symbol `__D3std6stream6Stream4readFJsZv'
socketstream.s:1416: Error: Local symbol `LTHUNK5' can't be equated to undefined symbol `__D3std6stream6Stream4readFJtZv'
socketstream.s:1416: Error: Local symbol `LTHUNK6' can't be equated to undefined symbol `__D3std6stream6Stream4readFJiZv'
socketstream.s:1416: Error: Local symbol `LTHUNK7' can't be equated to undefined symbol `__D3std6stream6Stream4readFJkZv'
socketstream.s:1416: Error: Local symbol `LTHUNK8' can't be equated to undefined symbol `__D3std6stream6Stream4readFJlZv'
socketstream.s:1416: Error: Local symbol `LTHUNK9' can't be equated to undefined symbol `__D3std6stream6Stream4readFJmZv'
socketstream.s:1416: Error: Local symbol `LTHUNK10' can't be equated to undefined symbol `__D3std6stream6Stream4readFJfZv'
socketstream.s:1416: Error: Local symbol `LTHUNK11' can't be equated to undefined symbol `__D3std6stream6Stream4readFJdZv'
socketstream.s:1416: Error: Local symbol `LTHUNK12' can't be equated to undefined symbol `__D3std6stream6Stream4readFJeZv'
socketstream.s:1416: Error: Local symbol `LTHUNK13' can't be equated to undefined symbol `__D3std6stream6Stream4readFJoZv'
socketstream.s:1416: Error: Local symbol `LTHUNK14' can't be equated to undefined symbol `__D3std6stream6Stream4readFJpZv'
socketstream.s:1416: Error: Local symbol `LTHUNK15' can't be equated to undefined symbol `__D3std6stream6Stream4readFJjZv'
socketstream.s:1416: Error: Local symbol `LTHUNK16' can't be equated to undefined symbol `__D3std6stream6Stream4readFJqZv'
socketstream.s:1416: Error: Local symbol `LTHUNK17' can't be equated to undefined symbol `__D3std6stream6Stream4readFJrZv'
socketstream.s:1416: Error: Local symbol `LTHUNK18' can't be equated to undefined symbol `__D3std6stream6Stream4readFJcZv'
socketstream.s:1416: Error: Local symbol `LTHUNK19' can't be equated to undefined symbol `__D3std6stream6Stream4readFJaZv'
socketstream.s:1416: Error: Local symbol `LTHUNK20' can't be equated to undefined symbol `__D3std6stream6Stream4readFJuZv'
socketstream.s:1416: Error: Local symbol `LTHUNK21' can't be equated to undefined symbol `__D3std6stream6Stream4readFJwZv'
socketstream.s:1416: Error: Local symbol `LTHUNK22' can't be equated to undefined symbol `__D3std6stream6Stream4readFJAaZv'
socketstream.s:1416: Error: Local symbol `LTHUNK23' can't be equated to undefined symbol `__D3std6stream6Stream4readFJAuZv'
socketstream.s:1416: Error: Local symbol `LTHUNK25' can't be equated to undefined symbol `__D3std6stream6Stream8readLineFAaZAa'
socketstream.s:1416: Error: Local symbol `LTHUNK27' can't be equated to undefined symbol `__D3std6stream6Stream9readLineWFAuZAu'
socketstream.s:1416: Error: Local symbol `LTHUNK28' can't be equated to undefined symbol `__D3std6stream6Stream7opApplyFDFKAaZiZi'
socketstream.s:1416: Error: Local symbol `LTHUNK29' can't be equated to undefined symbol `__D3std6stream6Stream7opApplyFDFKmKAaZiZi'
socketstream.s:1416: Error: Local symbol `LTHUNK30' can't be equated to undefined symbol `__D3std6stream6Stream7opApplyFDFKAuZiZi'
socketstream.s:1416: Error: Local symbol `LTHUNK31' can't be equated to undefined symbol `__D3std6stream6Stream7opApplyFDFKmKAuZiZi'
socketstream.s:1416: Error: Local symbol `LTHUNK32' can't be equated to undefined symbol `__D3std6stream6Stream10readStringFkZAa'
socketstream.s:1416: Error: Local symbol `LTHUNK33' can't be equated to undefined symbol `__D3std6stream6Stream11readStringWFkZAu'
socketstream.s:1416: Error: Local symbol `LTHUNK36' can't be equated to undefined symbol `__D3std6stream6Stream6ungetcFaZa'
socketstream.s:1416: Error: Local symbol `LTHUNK37' can't be equated to undefined symbol `__D3std6stream6Stream7ungetcwFuZu'
socketstream.s:1416: Error: Local symbol `LTHUNK38' can't be equated to undefined symbol `__D3std6stream6Stream6vscanfFAaPaZi'
socketstream.s:1416: Error: Local symbol `LTHUNK39' can't be equated to undefined symbol `__D3std6stream6Stream5scanfFAaYi'
socketstream.s:1416: Error: Local symbol `LTHUNK40' can't be equated to undefined symbol `__D3std6stream6Stream9availableFZk'
socketstream.s:1416: Error: Local symbol `LTHUNK42' can't be equated to undefined symbol `__D3std6stream6Stream6isOpenFZb'
socketstream.s:1416: Error: Local symbol `LTHUNK43' can't be equated to undefined symbol `__D3std6stream6Stream10writeExactFPvkZv'
socketstream.s:1416: Error: Local symbol `LTHUNK44' can't be equated to undefined symbol `__D3std6stream6Stream5writeFAhZk'
socketstream.s:1416: Error: Local symbol `LTHUNK45' can't be equated to undefined symbol `__D3std6stream6Stream5writeFgZv'
socketstream.s:1416: Error: Local symbol `LTHUNK46' can't be equated to undefined symbol `__D3std6stream6Stream5writeFhZv'
socketstream.s:1416: Error: Local symbol `LTHUNK47' can't be equated to undefined symbol `__D3std6stream6Stream5writeFsZv'
socketstream.s:1416: Error: Local symbol `LTHUNK48' can't be equated to undefined symbol `__D3std6stream6Stream5writeFtZv'
socketstream.s:1416: Error: Local symbol `LTHUNK49' can't be equated to undefined symbol `__D3std6stream6Stream5writeFiZv'
socketstream.s:1416: Error: Local symbol `LTHUNK50' can't be equated to undefined symbol `__D3std6stream6Stream5writeFkZv'
socketstream.s:1416: Error: Local symbol `LTHUNK51' can't be equated to undefined symbol `__D3std6stream6Stream5writeFlZv'
socketstream.s:1416: Error: Local symbol `LTHUNK52' can't be equated to undefined symbol `__D3std6stream6Stream5writeFmZv'
socketstream.s:1416: Error: Local symbol `LTHUNK53' can't be equated to undefined symbol `__D3std6stream6Stream5writeFfZv'
socketstream.s:1416: Error: Local symbol `LTHUNK54' can't be equated to undefined symbol `__D3std6stream6Stream5writeFdZv'
socketstream.s:1416: Error: Local symbol `LTHUNK55' can't be equated to undefined symbol `__D3std6stream6Stream5writeFeZv'
socketstream.s:1416: Error: Local symbol `LTHUNK56' can't be equated to undefined symbol `__D3std6stream6Stream5writeFoZv'
socketstream.s:1416: Error: Local symbol `LTHUNK57' can't be equated to undefined symbol `__D3std6stream6Stream5writeFpZv'
socketstream.s:1416: Error: Local symbol `LTHUNK58' can't be equated to undefined symbol `__D3std6stream6Stream5writeFjZv'
socketstream.s:1416: Error: Local symbol `LTHUNK59' can't be equated to undefined symbol `__D3std6stream6Stream5writeFqZv'
socketstream.s:1416: Error: Local symbol `LTHUNK60' can't be equated to undefined symbol `__D3std6stream6Stream5writeFrZv'
socketstream.s:1416: Error: Local symbol `LTHUNK61' can't be equated to undefined symbol `__D3std6stream6Stream5writeFcZv'
socketstream.s:1416: Error: Local symbol `LTHUNK62' can't be equated to undefined symbol `__D3std6stream6Stream5writeFaZv'
socketstream.s:1416: Error: Local symbol `LTHUNK63' can't be equated to undefined symbol `__D3std6stream6Stream5writeFuZv'
socketstream.s:1416: Error: Local symbol `LTHUNK64' can't be equated to undefined symbol `__D3std6stream6Stream5writeFwZv'
socketstream.s:1416: Error: Local symbol `LTHUNK65' can't be equated to undefined symbol `__D3std6stream6Stream5writeFAaZv'
socketstream.s:1416: Error: Local symbol `LTHUNK66' can't be equated to undefined symbol `__D3std6stream6Stream5writeFAuZv'
socketstream.s:1416: Error: Local symbol `LTHUNK67' can't be equated to undefined symbol `__D3std6stream6Stream9writeLineFAaZv'
socketstream.s:1416: Error: Local symbol `LTHUNK68' can't be equated to undefined symbol `__D3std6stream6Stream10writeLineWFAuZv'
socketstream.s:1416: Error: Local symbol `LTHUNK69' can't be equated to undefined symbol `__D3std6stream6Stream11writeStringFAaZv'
socketstream.s:1416: Error: Local symbol `LTHUNK70' can't be equated to undefined symbol `__D3std6stream6Stream12writeStringWFAuZv'
socketstream.s:1416: Error: Local symbol `LTHUNK71' can't be equated to undefined symbol `__D3std6stream6Stream7vprintfFAaPaZk'
socketstream.s:1416: Error: Local symbol `LTHUNK72' can't be equated to undefined symbol `__D3std6stream6Stream6printfFAaYk'
socketstream.s:1416: Error: Local symbol `LTHUNK73' can't be equated to undefined symbol `__D3std6stream6Stream6writefFYC3std6stream6Stream'
socketstream.s:1416: Error: Local symbol `LTHUNK74' can't be equated to undefined symbol `__D3std6stream6Stream8writeflnFYC3std6stream6Stream'
socketstream.s:1416: Error: Local symbol `LTHUNK75' can't be equated to undefined symbol `__D3std6stream6Stream5flushFZv'
socketstream.s:1416: Error: Local symbol `LTHUNK77' can't be equated to undefined symbol `__D3std6stream6Stream6isOpenFZb'


Gerrit
-- 
=^..^=
June 07, 2005
Gerrit P. Haase wrote:
> Hi all,
> 
> I'm trying to integrate the D frontend for GCC in the GCC 3.4.4 release,
> I'm in contact with the maintainer David Friedman, he successfully
> compiled the current (0.12.1) sources with MinGW and I expect that he
> also has tested it with other supported platforms, however I get this
> error now and I have not really an idea what the reason may be.  I found
> one patch for gas for another platform:
> http://sources.redhat.com/bugzilla/show_bug.cgi?id=857
> http://sources.redhat.com/bugzilla/show_bug.cgi?id=936
> http://sourceware.org/ml/binutils/2005-05/msg00289.html
> 
> However I don't know if this is related or if there may be another problem with the e.g. source.
> 
> What is new in the source regarding this problem is this patch which gives me the error cited below:

> ***************
> *** 1787,1799 ****
>   IRState::outputThunk(tree thunk_decl, tree target_decl, int offset)
>   {
>       int delta = -offset;
> !     tree alias = target_decl;
> 
>       TREE_ADDRESSABLE(target_decl) = 1;
>       TREE_USED(target_decl) = 1;
>   #if D_GCC_VER >= 34
>       DECL_VISIBILITY (thunk_decl) = DECL_VISIBILITY (target_decl);
> ! #ifdef ASM_OUTPUT_DEF
>     if (targetm.have_named_sections)
>       {
>         resolve_unique_section (target_decl, 0, flag_function_sections);
> --- 1844,1866 ----
>   IRState::outputThunk(tree thunk_decl, tree target_decl, int offset)
>   {
>       int delta = -offset;
> !     tree alias;
> ! #if D_GCC_VER >= 34
> ! #ifdef ASM_OUTPUT_DEF
> !     alias = make_alias_for_thunk(target_decl);
> ! #else
> !     alias = target_decl;
> ! #endif
> ! #else
> !     alias = target_decl;
> ! #endif
> 
>       TREE_ADDRESSABLE(target_decl) = 1;
>       TREE_USED(target_decl) = 1;
>   #if D_GCC_VER >= 34
>       DECL_VISIBILITY (thunk_decl) = DECL_VISIBILITY (target_decl);
> ! #if defined (ASM_OUTPUT_DEF) \
> !   && !defined (TARGET_IS_PE_COFF)
>     if (targetm.have_named_sections)
>       {
>         resolve_unique_section (target_decl, 0, flag_function_sections);

Maybe this is the problem here:
!   && !defined (TARGET_IS_PE_COFF)


Is TARGET_IS_PE_COFF defined for Cygwin?


Gerrit
-- 
=^..^=
June 07, 2005
Gerrit P. Haase wrote:

> Gerrit P. Haase wrote:
> 
>> Hi all,
>>
>> I'm trying to integrate the D frontend for GCC in the GCC 3.4.4 release,
>> I'm in contact with the maintainer David Friedman, he successfully
>> compiled the current (0.12.1) sources with MinGW and I expect that he
>> also has tested it with other supported platforms, however I get this
>> error now and I have not really an idea what the reason may be.  I found
>> one patch for gas for another platform:
>> http://sources.redhat.com/bugzilla/show_bug.cgi?id=857
>> http://sources.redhat.com/bugzilla/show_bug.cgi?id=936
>> http://sourceware.org/ml/binutils/2005-05/msg00289.html
>>
>> However I don't know if this is related or if there may be another problem with the e.g. source.
>>
>> What is new in the source regarding this problem is this patch which gives me the error cited below:
> 
> 
>> ***************
>> *** 1787,1799 ****
>>   IRState::outputThunk(tree thunk_decl, tree target_decl, int offset)
>>   {
>>       int delta = -offset;
>> !     tree alias = target_decl;
>>
>>       TREE_ADDRESSABLE(target_decl) = 1;
>>       TREE_USED(target_decl) = 1;
>>   #if D_GCC_VER >= 34
>>       DECL_VISIBILITY (thunk_decl) = DECL_VISIBILITY (target_decl);
>> ! #ifdef ASM_OUTPUT_DEF
>>     if (targetm.have_named_sections)
>>       {
>>         resolve_unique_section (target_decl, 0, flag_function_sections);
>> --- 1844,1866 ----
>>   IRState::outputThunk(tree thunk_decl, tree target_decl, int offset)
>>   {
>>       int delta = -offset;
>> !     tree alias;
>> ! #if D_GCC_VER >= 34
>> ! #ifdef ASM_OUTPUT_DEF
>> !     alias = make_alias_for_thunk(target_decl);
>> ! #else
>> !     alias = target_decl;
>> ! #endif
>> ! #else
>> !     alias = target_decl;
>> ! #endif
>>
>>       TREE_ADDRESSABLE(target_decl) = 1;
>>       TREE_USED(target_decl) = 1;
>>   #if D_GCC_VER >= 34
>>       DECL_VISIBILITY (thunk_decl) = DECL_VISIBILITY (target_decl);
>> ! #if defined (ASM_OUTPUT_DEF) \
>> !   && !defined (TARGET_IS_PE_COFF)
>>     if (targetm.have_named_sections)
>>       {
>>         resolve_unique_section (target_decl, 0, flag_function_sections);
> 
> 
> Maybe this is the problem here:
> !   && !defined (TARGET_IS_PE_COFF)
> 
> 
> Is TARGET_IS_PE_COFF defined for Cygwin?

Yes it is:

#define TARGET_IS_PE_COFF 1

Is this wrong in the d-codegen source?  I'kll try what changes if I include Cygwin:
! #if defined (ASM_OUTPUT_DEF) \
!   && (!defined (TARGET_IS_PE_COFF) && !defined(__CYGWIN__))


Gerrit
June 07, 2005
Gerrit P. Haase wrote:
> Gerrit P. Haase wrote:
> 
>> Gerrit P. Haase wrote:
>>
>>> Hi all,
>>>
>>> I'm trying to integrate the D frontend for GCC in the GCC 3.4.4 release,
>>> I'm in contact with the maintainer David Friedman, he successfully
>>> compiled the current (0.12.1) sources with MinGW and I expect that he
>>> also has tested it with other supported platforms, however I get this
>>> error now and I have not really an idea what the reason may be.  I found
>>> one patch for gas for another platform:
>>> http://sources.redhat.com/bugzilla/show_bug.cgi?id=857
>>> http://sources.redhat.com/bugzilla/show_bug.cgi?id=936
>>> http://sourceware.org/ml/binutils/2005-05/msg00289.html
>>>
>>> However I don't know if this is related or if there may be another problem with the e.g. source.
>>>
>>> What is new in the source regarding this problem is this patch which gives me the error cited below:
>>
>>
>>
>>> ***************
>>> *** 1787,1799 ****
>>>   IRState::outputThunk(tree thunk_decl, tree target_decl, int offset)
>>>   {
>>>       int delta = -offset;
>>> !     tree alias = target_decl;
>>>
>>>       TREE_ADDRESSABLE(target_decl) = 1;
>>>       TREE_USED(target_decl) = 1;
>>>   #if D_GCC_VER >= 34
>>>       DECL_VISIBILITY (thunk_decl) = DECL_VISIBILITY (target_decl);
>>> ! #ifdef ASM_OUTPUT_DEF
>>>     if (targetm.have_named_sections)
>>>       {
>>>         resolve_unique_section (target_decl, 0, flag_function_sections);
>>> --- 1844,1866 ----
>>>   IRState::outputThunk(tree thunk_decl, tree target_decl, int offset)
>>>   {
>>>       int delta = -offset;
>>> !     tree alias;
>>> ! #if D_GCC_VER >= 34
>>> ! #ifdef ASM_OUTPUT_DEF
>>> !     alias = make_alias_for_thunk(target_decl);
>>> ! #else
>>> !     alias = target_decl;
>>> ! #endif
>>> ! #else
>>> !     alias = target_decl;
>>> ! #endif
>>>
>>>       TREE_ADDRESSABLE(target_decl) = 1;
>>>       TREE_USED(target_decl) = 1;
>>>   #if D_GCC_VER >= 34
>>>       DECL_VISIBILITY (thunk_decl) = DECL_VISIBILITY (target_decl);
>>> ! #if defined (ASM_OUTPUT_DEF) \
>>> !   && !defined (TARGET_IS_PE_COFF)
>>>     if (targetm.have_named_sections)
>>>       {
>>>         resolve_unique_section (target_decl, 0, flag_function_sections);
>>
>>
>>
>> Maybe this is the problem here:
>> !   && !defined (TARGET_IS_PE_COFF)
>>
>>
>> Is TARGET_IS_PE_COFF defined for Cygwin?
> 
> 
> Yes it is:
> 
> #define TARGET_IS_PE_COFF 1
> 
> Is this wrong in the d-codegen source?  I'kll try what changes if I include Cygwin:
> ! #if defined (ASM_OUTPUT_DEF) \
> !   && (!defined (TARGET_IS_PE_COFF) && !defined(__CYGWIN__))
> 
> 
> Gerrit

That check for TARGET_IS_PE_COFF is necessary.  Without it, there is a very nasty code gen bug that won't show up until you run the code.

This is probably a better test:

--- d-codegen.cc.orig	Tue Jun  7 14:10:57 2005
+++ d-codegen.cc	Tue Jun  7 14:11:55 2005
@@ -1757,7 +1757,7 @@
   char buf[256];

 #if defined (TARGET_IS_PE_COFF)
-  if (DECL_ONE_ONLY (function))
+  //  if (DECL_ONE_ONLY (function))
     return function;
 #endif

It might break -fpic code generation again, though.  I won't be able to test it until later today, but you can try it yourself by adding "-fpic" to the DFLAGS in Phobos Makefile.

David

June 07, 2005
Gerrit P. Haase wrote:

>> Maybe this is the problem here:
>> !   && !defined (TARGET_IS_PE_COFF)
>>
>>
>> Is TARGET_IS_PE_COFF defined for Cygwin?
> 
> 
> Yes it is:
> 
> #define TARGET_IS_PE_COFF 1
> 
> Is this wrong in the d-codegen source?  I'kll try what changes if I include Cygwin:
> ! #if defined (ASM_OUTPUT_DEF) \
> !   && (!defined (TARGET_IS_PE_COFF) && !defined(__CYGWIN__))

Nope, this was not the problem.


Gerrit
June 07, 2005
Gerrit P. Haase wrote:

> Hi all,
> 
> I'm trying to integrate the D frontend for GCC in the GCC 3.4.4 release,
> I'm in contact with the maintainer David Friedman, he successfully
> compiled the current (0.12.1) sources with MinGW and I expect that he
> also has tested it with other supported platforms, however I get this
> error now and I have not really an idea what the reason may be. 

[...]

> What is new in the source regarding this problem is this patch which gives me the error cited below:

[...]

This part is 1:1 from C++/methods.c, besides the commented lines, these are ncluded in the C++ compiler, so it seems that one or more of these
comments here are too much because the C++ compiler is working:

> diff -cr d-0.12/d-codegen.cc d-0.12.1/d-codegen.cc
> *** d-0.12/d-codegen.cc    Sat May 28 18:00:56 2005
> --- d-0.12.1/d-codegen.cc    Sun Jun  5 19:27:36 2005
> ***************
> *** 1743,1748 ****
> --- 1743,1805 ----
>       return call_exp;
>   }
> 
> +
> + #if D_GCC_VER >= 34
> + #ifdef ASM_OUTPUT_DEF
> + static int thunk_labelno;
> +
> + /* Create a static alias to function.  */
> +
> + static tree
> + make_alias_for_thunk (tree function)
> + {
> +   tree alias;
> +   char buf[256];
> +
> + #if defined (TARGET_IS_PE_COFF)
> +   if (DECL_ONE_ONLY (function))
> +     return function;
> + #endif
> +
> +   ASM_GENERATE_INTERNAL_LABEL (buf, "LTHUNK", thunk_labelno);
> +   thunk_labelno++;
> +   alias = build_decl (FUNCTION_DECL, get_identifier (buf),
> +               TREE_TYPE (function));
> +   //DECL_LANG_SPECIFIC (alias) = DECL_LANG_SPECIFIC (function);
> +   //cxx_dup_lang_specific_decl (alias);
> +   DECL_CONTEXT (alias) = NULL;
> +   TREE_READONLY (alias) = TREE_READONLY (function);
> +   TREE_THIS_VOLATILE (alias) = TREE_THIS_VOLATILE (function);
> +   TREE_PUBLIC (alias) = 0;
> +   //DECL_INTERFACE_KNOWN (alias) = 1;
> +   //DECL_NOT_REALLY_EXTERN (alias) = 1;
> +   //DECL_THIS_STATIC (alias) = 1;
> +   //DECL_SAVED_FUNCTION_DATA (alias) = NULL;
> +   //DECL_DESTRUCTOR_P (alias) = 0;
> +   //DECL_CONSTRUCTOR_P (alias) = 0;
> +   //DECL_CLONED_FUNCTION (alias) = NULL_TREE;
> +   DECL_EXTERNAL (alias) = 0;
> +   DECL_ARTIFICIAL (alias) = 1;
> +   DECL_NO_STATIC_CHAIN (alias) = 1;
> +   //DECL_PENDING_INLINE_P (alias) = 0;
> +   DECL_INLINE (alias) = 0;
> +   DECL_DECLARED_INLINE_P (alias) = 0;
> +   //DECL_DEFERRED_FN (alias) = 0;
> +   //DECL_USE_TEMPLATE (alias) = 0;
> +   //DECL_TEMPLATE_INSTANTIATED (alias) = 0;
> +   //DECL_TEMPLATE_INFO (alias) = NULL;
> +   //DECL_INITIAL (alias) = error_mark_node;
> +   TREE_ADDRESSABLE (alias) = 1;
> +   TREE_USED (alias) = 1;
> +   SET_DECL_ASSEMBLER_NAME (alias, DECL_NAME (alias));
> +   TREE_SYMBOL_REFERENCED (DECL_ASSEMBLER_NAME (alias)) = 1;
> +   if (!flag_syntax_only)
> +     assemble_alias (alias, DECL_ASSEMBLER_NAME (function));
> +   return alias;
> + }
> + #endif
> + #endif
> +
>   // Can't output thunks while a function is being compiled. Called from
>   // d-decls.cc (no cur_irs)
>   static Array deferredThunks;


Gerrit
June 07, 2005
David Friedman wrote:
>>> Maybe this is the problem here:
>>> !   && !defined (TARGET_IS_PE_COFF)
>>>
>>>
>>> Is TARGET_IS_PE_COFF defined for Cygwin?
>>
>>
>>
>> Yes it is:
>>
>> #define TARGET_IS_PE_COFF 1
>>
>> Is this wrong in the d-codegen source?  I'kll try what changes if I include Cygwin:
>> ! #if defined (ASM_OUTPUT_DEF) \
>> !   && (!defined (TARGET_IS_PE_COFF) && !defined(__CYGWIN__))
>>
>>
>> Gerrit
> 
> 
> That check for TARGET_IS_PE_COFF is necessary.  Without it, there is a very nasty code gen bug that won't show up until you run the code.
> 
> This is probably a better test:
> 
> --- d-codegen.cc.orig    Tue Jun  7 14:10:57 2005
> +++ d-codegen.cc    Tue Jun  7 14:11:55 2005
> @@ -1757,7 +1757,7 @@
>    char buf[256];
> 
>  #if defined (TARGET_IS_PE_COFF)
> -  if (DECL_ONE_ONLY (function))
> +  //  if (DECL_ONE_ONLY (function))
>      return function;
>  #endif
> 
> It might break -fpic code generation again, though.  I won't be able to test it until later today, but you can try it yourself by adding "-fpic" to the DFLAGS in Phobos Makefile.

-fpic is totally unsupported on Windows because on Windows everything is
position independat anyway.  I'll try this now.


See you later,
Gerrit
June 07, 2005
David Friedman wrote:
> This is probably a better test:
> 
> --- d-codegen.cc.orig    Tue Jun  7 14:10:57 2005
> +++ d-codegen.cc    Tue Jun  7 14:11:55 2005
> @@ -1757,7 +1757,7 @@
>    char buf[256];
> 
>  #if defined (TARGET_IS_PE_COFF)
> -  if (DECL_ONE_ONLY (function))
> +  //  if (DECL_ONE_ONLY (function))
>      return function;
>  #endif
> 
> It might break -fpic code generation again, though.  I won't be able to test it until later today, but you can try it yourself by adding "-fpic" to the DFLAGS in Phobos Makefile.

Yes, now it compiles, thank you.


Gerrit