unresolved symbols during binutils linking

unresolved symbols during binutils linking

[Richard Herveille]

Hi,

I am trying to port binutils to a new CPU using CGEN.
My approach ... start with a basic/simple .cpu file (with limited 
opcodes) and add instructions once it all starts working.

I think I modified all configure/makefiles in binutils; at least it is 
alive :-)
I can also see that cgen gets called and generates the files without 
issuing any errors. I can also see the opcode definitions in teh fiels.
However I am getting errors during compilation (actually during linking);

Example:
libtool: link: gcc -W ... ../opcodes/.libs/libopcodes.a(mycpu-dis.o): In 
function 'mycpu_cgen_init_dis':
... mycpu-dis.c:148 undefined reference to 'mycpu_cgen_init_ibld_table'

other undefined references:
mycpu_cgen_cpu_open, mycpu_cgen_opval_h_gpr, mycpu_cgen_opval_h_gpar, 
mycpu_cgen_init_opcode_table, mycpu_cgen_build_insn_regex

I can see all of these functions being defined in the generated C files. 
I can also see that the names are present in libopcodes.a.

Does anybody have any clue what is going wrong? I tried other targets 
and they work out of the box. So it must be something in my CPU's setup.

Thanks,
Richard

Re: unresolved symbols during binutils linking

[Jean-Marc Saffroy]

Hi Richard,

Maybe you forgot to add mycpu-{asm,desc,dis,ibld,opc}.c to 
TARGET_LIBOPCODES_CFILES in opcodes/Makefile.am (and of course 
regenerate Makefile.in and rerun configure) ?

Cheers,
JM

Richard Herveille wrote:
> Hi,
> 
> I am trying to port binutils to a new CPU using CGEN.
> My approach ... start with a basic/simple .cpu file (with limited 
> opcodes) and add instructions once it all starts working.
> 
> I think I modified all configure/makefiles in binutils; at least it is 
> alive :-)
> I can also see that cgen gets called and generates the files without 
> issuing any errors. I can also see the opcode definitions in teh fiels.
> However I am getting errors during compilation (actually during linking);
> 
> Example:
> libtool: link: gcc -W ... ../opcodes/.libs/libopcodes.a(mycpu-dis.o): In 
> function 'mycpu_cgen_init_dis':
> ... mycpu-dis.c:148 undefined reference to 'mycpu_cgen_init_ibld_table'
> 
> other undefined references:
> mycpu_cgen_cpu_open, mycpu_cgen_opval_h_gpr, mycpu_cgen_opval_h_gpar, 
> mycpu_cgen_init_opcode_table, mycpu_cgen_build_insn_regex
> 
> I can see all of these functions being defined in the generated C files. 
> I can also see that the names are present in libopcodes.a.
> 
> Does anybody have any clue what is going wrong? I tried other targets 
> and they work out of the box. So it must be something in my CPU's setup.
> 
> Thanks,
> Richard
> 
> 
> 

Re: unresolved symbols during binutils linking

[Jean-Marc Saffroy]

Hmm it seems there are other places where you need to mention these files.
FWIW below is the patch I use.

HTH,
JM

Jean-Marc Saffroy wrote:
> Hi Richard,
> 
> Maybe you forgot to add mycpu-{asm,desc,dis,ibld,opc}.c to
> TARGET_LIBOPCODES_CFILES in opcodes/Makefile.am (and of course
> regenerate Makefile.in and rerun configure) ?
> 
> Cheers,
> JM
> 
> Richard Herveille wrote:
>> Hi,
>>
>> I am trying to port binutils to a new CPU using CGEN.
>> My approach ... start with a basic/simple .cpu file (with limited
>> opcodes) and add instructions once it all starts working.
>>
>> I think I modified all configure/makefiles in binutils; at least it is
>> alive :-)
>> I can also see that cgen gets called and generates the files without
>> issuing any errors. I can also see the opcode definitions in teh fiels.
>> However I am getting errors during compilation (actually during linking);
>>
>> Example:
>> libtool: link: gcc -W ... ../opcodes/.libs/libopcodes.a(mycpu-dis.o):
>> In function 'mycpu_cgen_init_dis':
>> ... mycpu-dis.c:148 undefined reference to 'mycpu_cgen_init_ibld_table'
>>
>> other undefined references:
>> mycpu_cgen_cpu_open, mycpu_cgen_opval_h_gpr, mycpu_cgen_opval_h_gpar,
>> mycpu_cgen_init_opcode_table, mycpu_cgen_build_insn_regex
>>
>> I can see all of these functions being defined in the generated C
>> files. I can also see that the names are present in libopcodes.a.
>>
>> Does anybody have any clue what is going wrong? I tried other targets
>> and they work out of the box. So it must be something in my CPU's setup.
>>
>> Thanks,
>> Richard
>>
>>
>>
> 
> 

--- a/include/dis-asm.h	Mon Mar 22 15:03:12 2010 +0100
+++ b/include/dis-asm.h	Mon Mar 22 16:02:04 2010 +0100
@@ -273,6 +273,7 @@
 extern int print_insn_msp430		(bfd_vma, disassemble_info *);
 extern int print_insn_mt                (bfd_vma, disassemble_info *);
 extern int print_insn_ns32k		(bfd_vma, disassemble_info *);
+extern int print_insn_mycpu		(bfd_vma, disassemble_info *);
 extern int print_insn_openrisc		(bfd_vma, disassemble_info *);
 extern int print_insn_pdp11		(bfd_vma, disassemble_info *);
 extern int print_insn_pj		(bfd_vma, disassemble_info *);
--- a/opcodes/Makefile.am	Mon Mar 22 15:03:12 2010 +0100
+++ b/opcodes/Makefile.am	Mon Mar 22 16:02:04 2010 +0100
@@ -174,6 +174,11 @@
 	mt-ibld.c \
 	mt-opc.c \
 	ns32k-dis.c \
+	mycpu-asm.c \
+	mycpu-desc.c \
+	mycpu-dis.c \
+	mycpu-ibld.c \
+	mycpu-opc.c \
 	openrisc-asm.c \
 	openrisc-desc.c \
 	openrisc-dis.c \
@@ -305,7 +310,7 @@
 	  && mv tmp $(srcdir)/po/POTFILES.in
 
 CLEANFILES = \
-	stamp-ip2k stamp-m32c stamp-m32r stamp-fr30 stamp-frv \
+	stamp-ip2k stamp-m32c stamp-m32r stamp-fr30 stamp-frv stamp-mycpu \
 	stamp-openrisc stamp-iq2000 stamp-mep stamp-mt stamp-xstormy16 stamp-xc16x\
 	libopcodes.a stamp-lib dep.sed DEP DEPA DEP1 DEP2
 
@@ -322,7 +327,7 @@
 	$(CGENDIR)/opc-opinst.scm \
 	cgen-asm.in cgen-dis.in cgen-ibld.in
 
-CGEN_CPUS = fr30 frv ip2k iq2000 lm32 m32c m32r mep mt openrisc xc16x xstormy16
+CGEN_CPUS = fr30 frv ip2k iq2000 lm32 m32c m32r mep mt mycpu openrisc xc16x xstormy16
 
 if CGEN_MAINT
 FR30_DEPS = stamp-fr30
@@ -334,6 +339,7 @@
 M32R_DEPS = stamp-m32r
 MEP_DEPS = stamp-mep
 MT_DEPS = stamp-mt
+MYCPU_DEPS = stamp-mycpu
 OPENRISC_DEPS = stamp-openrisc
 XC16X_DEPS = stamp-xc16x
 XSTORMY16_DEPS = stamp-xstormy16
@@ -347,6 +353,7 @@
 M32R_DEPS =
 MEP_DEPS =
 MT_DEPS =
+MYCPU_DEPS =
 OPENRISC_DEPS = 
 XC16X_DEPS = 
 XSTORMY16_DEPS = 
@@ -431,6 +438,12 @@
 		archfile=$(srcdir)/../cpu/mt.cpu \
 		opcfile=$(srcdir)/../cpu/mt.opc extrafiles=
 
+$(srcdir)/mycpu-desc.h $(srcdir)/mycpu-desc.c $(srcdir)/mycpu-opc.h $(srcdir)/mycpu-opc.c $(srcdir)/mycpu-ibld.c $(srcdir)/mycpu-asm.c $(srcdir)/mycpu-dis.c: $(MYCPU_DEPS)
+	@true
+stamp-mycpu: $(CGENDEPS) $(srcdir)/../cpu/mycpu.cpu $(srcdir)/../cpu/mycpu.opc
+	$(MAKE) run-cgen arch=mycpu prefix=mycpu options= \
+		 archfile=$(srcdir)/../cpu/mycpu.cpu opcfile=$(srcdir)/../cpu/mycpu.opc extrafiles=
+
 $(srcdir)/openrisc-desc.h $(srcdir)/openrisc-desc.c $(srcdir)/openrisc-opc.h $(srcdir)/openrisc-opc.c $(srcdir)/openrisc-ibld.c $(srcdir)/openrisc-asm.c $(srcdir)/openrisc-dis.c: $(OPENRISC_DEPS)
 	@true
 stamp-openrisc: $(CGENDEPS) $(CPUDIR)/openrisc.cpu $(CPUDIR)/openrisc.opc
--- a/opcodes/configure.in	Mon Mar 22 15:03:12 2010 +0100
+++ b/opcodes/configure.in	Mon Mar 22 16:02:04 2010 +0100
@@ -262,6 +262,7 @@
 	bfd_mt_arch)		ta="$ta mt-asm.lo mt-desc.lo mt-dis.lo mt-ibld.lo mt-opc.lo" using_cgen=yes ;;
 	bfd_msp430_arch)	ta="$ta msp430-dis.lo" ;;
 	bfd_ns32k_arch)		ta="$ta ns32k-dis.lo" ;;
+	bfd_mycpu_arch)	ta="$ta mycpu-asm.lo mycpu-desc.lo mycpu-dis.lo mycpu-ibld.lo mycpu-opc.lo" using_cgen=yes ;;
 	bfd_openrisc_arch)	ta="$ta openrisc-asm.lo openrisc-desc.lo openrisc-dis.lo openrisc-ibld.lo openrisc-opc.lo" using_cgen=yes ;;
 	bfd_or32_arch)		ta="$ta or32-dis.lo or32-opc.lo" using_cgen=yes ;;
 	bfd_pdp11_arch)		ta="$ta pdp11-dis.lo pdp11-opc.lo" ;;
--- a/opcodes/disassemble.c	Mon Mar 22 15:03:12 2010 +0100
+++ b/opcodes/disassemble.c	Mon Mar 22 16:02:04 2010 +0100
@@ -65,6 +65,7 @@
 #define ARCH_mt
 #define ARCH_msp430
 #define ARCH_ns32k
+#define ARCH_mycpu
 #define ARCH_openrisc
 #define ARCH_or32
 #define ARCH_pdp11
@@ -319,6 +320,11 @@
       disassemble = print_insn_openrisc;
       break;
 #endif
+#ifdef ARCH_mycpu
+    case bfd_arch_mycpu:
+      disassemble = print_insn_mycpu;
+      break;
+#endif
 #ifdef ARCH_or32
     case bfd_arch_or32:
       if (bfd_big_endian (abfd))

Re: unresolved symbols during binutils linking

[Richard Herveille]

I couldn't figure it out...

However I got it to work, at least pass this issue.
I downloaded the latest cgen snapshot. This caught one issue in my .cpu 
file. Though functional only.
I also downloaded the latest binutils snapshot and started from scratch :-s

That (somehow) fixed it.

Thanks for your help.
Richard

> Hmm it seems there are other places where you need to mention these files.
> FWIW below is the patch I use.
>
> HTH,
> JM
>
> Jean-Marc Saffroy wrote:
>   
>> Hi Richard,
>>
>> Maybe you forgot to add mycpu-{asm,desc,dis,ibld,opc}.c to
>> TARGET_LIBOPCODES_CFILES in opcodes/Makefile.am (and of course
>> regenerate Makefile.in and rerun configure) ?
>>
>> Cheers,
>> JM
>>
>> Richard Herveille wrote:
>>     
>>> Hi,
>>>
>>> I am trying to port binutils to a new CPU using CGEN.
>>> My approach ... start with a basic/simple .cpu file (with limited
>>> opcodes) and add instructions once it all starts working.
>>>
>>> I think I modified all configure/makefiles in binutils; at least it is
>>> alive :-)
>>> I can also see that cgen gets called and generates the files without
>>> issuing any errors. I can also see the opcode definitions in teh fiels.
>>> However I am getting errors during compilation (actually during linking);
>>>
>>> Example:
>>> libtool: link: gcc -W ... ../opcodes/.libs/libopcodes.a(mycpu-dis.o):
>>> In function 'mycpu_cgen_init_dis':
>>> ... mycpu-dis.c:148 undefined reference to 'mycpu_cgen_init_ibld_table'
>>>
>>> other undefined references:
>>> mycpu_cgen_cpu_open, mycpu_cgen_opval_h_gpr, mycpu_cgen_opval_h_gpar,
>>> mycpu_cgen_init_opcode_table, mycpu_cgen_build_insn_regex
>>>
>>> I can see all of these functions being defined in the generated C
>>> files. I can also see that the names are present in libopcodes.a.
>>>
>>> Does anybody have any clue what is going wrong? I tried other targets
>>> and they work out of the box. So it must be something in my CPU's setup.
>>>
>>> Thanks,
>>> Richard
>>>
>>>
>>>
>>>       
>>     
>
> --- a/include/dis-asm.h	Mon Mar 22 15:03:12 2010 +0100
> +++ b/include/dis-asm.h	Mon Mar 22 16:02:04 2010 +0100
> @@ -273,6 +273,7 @@
>  extern int print_insn_msp430		(bfd_vma, disassemble_info *);
>  extern int print_insn_mt                (bfd_vma, disassemble_info *);
>  extern int print_insn_ns32k		(bfd_vma, disassemble_info *);
> +extern int print_insn_mycpu		(bfd_vma, disassemble_info *);
>  extern int print_insn_openrisc		(bfd_vma, disassemble_info *);
>  extern int print_insn_pdp11		(bfd_vma, disassemble_info *);
>  extern int print_insn_pj		(bfd_vma, disassemble_info *);
> --- a/opcodes/Makefile.am	Mon Mar 22 15:03:12 2010 +0100
> +++ b/opcodes/Makefile.am	Mon Mar 22 16:02:04 2010 +0100
> @@ -174,6 +174,11 @@
>  	mt-ibld.c \
>  	mt-opc.c \
>  	ns32k-dis.c \
> +	mycpu-asm.c \
> +	mycpu-desc.c \
> +	mycpu-dis.c \
> +	mycpu-ibld.c \
> +	mycpu-opc.c \
>  	openrisc-asm.c \
>  	openrisc-desc.c \
>  	openrisc-dis.c \
> @@ -305,7 +310,7 @@
>  	  && mv tmp $(srcdir)/po/POTFILES.in
>  
>  CLEANFILES = \
> -	stamp-ip2k stamp-m32c stamp-m32r stamp-fr30 stamp-frv \
> +	stamp-ip2k stamp-m32c stamp-m32r stamp-fr30 stamp-frv stamp-mycpu \
>  	stamp-openrisc stamp-iq2000 stamp-mep stamp-mt stamp-xstormy16 stamp-xc16x\
>  	libopcodes.a stamp-lib dep.sed DEP DEPA DEP1 DEP2
>  
> @@ -322,7 +327,7 @@
>  	$(CGENDIR)/opc-opinst.scm \
>  	cgen-asm.in cgen-dis.in cgen-ibld.in
>  
> -CGEN_CPUS = fr30 frv ip2k iq2000 lm32 m32c m32r mep mt openrisc xc16x xstormy16
> +CGEN_CPUS = fr30 frv ip2k iq2000 lm32 m32c m32r mep mt mycpu openrisc xc16x xstormy16
>  
>  if CGEN_MAINT
>  FR30_DEPS = stamp-fr30
> @@ -334,6 +339,7 @@
>  M32R_DEPS = stamp-m32r
>  MEP_DEPS = stamp-mep
>  MT_DEPS = stamp-mt
> +MYCPU_DEPS = stamp-mycpu
>  OPENRISC_DEPS = stamp-openrisc
>  XC16X_DEPS = stamp-xc16x
>  XSTORMY16_DEPS = stamp-xstormy16
> @@ -347,6 +353,7 @@
>  M32R_DEPS =
>  MEP_DEPS =
>  MT_DEPS =
> +MYCPU_DEPS =
>  OPENRISC_DEPS = 
>  XC16X_DEPS = 
>  XSTORMY16_DEPS = 
> @@ -431,6 +438,12 @@
>  		archfile=$(srcdir)/../cpu/mt.cpu \
>  		opcfile=$(srcdir)/../cpu/mt.opc extrafiles=
>  
> +$(srcdir)/mycpu-desc.h $(srcdir)/mycpu-desc.c $(srcdir)/mycpu-opc.h $(srcdir)/mycpu-opc.c $(srcdir)/mycpu-ibld.c $(srcdir)/mycpu-asm.c $(srcdir)/mycpu-dis.c: $(MYCPU_DEPS)
> +	@true
> +stamp-mycpu: $(CGENDEPS) $(srcdir)/../cpu/mycpu.cpu $(srcdir)/../cpu/mycpu.opc
> +	$(MAKE) run-cgen arch=mycpu prefix=mycpu options= \
> +		 archfile=$(srcdir)/../cpu/mycpu.cpu opcfile=$(srcdir)/../cpu/mycpu.opc extrafiles=
> +
>  $(srcdir)/openrisc-desc.h $(srcdir)/openrisc-desc.c $(srcdir)/openrisc-opc.h $(srcdir)/openrisc-opc.c $(srcdir)/openrisc-ibld.c $(srcdir)/openrisc-asm.c $(srcdir)/openrisc-dis.c: $(OPENRISC_DEPS)
>  	@true
>  stamp-openrisc: $(CGENDEPS) $(CPUDIR)/openrisc.cpu $(CPUDIR)/openrisc.opc
> --- a/opcodes/configure.in	Mon Mar 22 15:03:12 2010 +0100
> +++ b/opcodes/configure.in	Mon Mar 22 16:02:04 2010 +0100
> @@ -262,6 +262,7 @@
>  	bfd_mt_arch)		ta="$ta mt-asm.lo mt-desc.lo mt-dis.lo mt-ibld.lo mt-opc.lo" using_cgen=yes ;;
>  	bfd_msp430_arch)	ta="$ta msp430-dis.lo" ;;
>  	bfd_ns32k_arch)		ta="$ta ns32k-dis.lo" ;;
> +	bfd_mycpu_arch)	ta="$ta mycpu-asm.lo mycpu-desc.lo mycpu-dis.lo mycpu-ibld.lo mycpu-opc.lo" using_cgen=yes ;;
>  	bfd_openrisc_arch)	ta="$ta openrisc-asm.lo openrisc-desc.lo openrisc-dis.lo openrisc-ibld.lo openrisc-opc.lo" using_cgen=yes ;;
>  	bfd_or32_arch)		ta="$ta or32-dis.lo or32-opc.lo" using_cgen=yes ;;
>  	bfd_pdp11_arch)		ta="$ta pdp11-dis.lo pdp11-opc.lo" ;;
> --- a/opcodes/disassemble.c	Mon Mar 22 15:03:12 2010 +0100
> +++ b/opcodes/disassemble.c	Mon Mar 22 16:02:04 2010 +0100
> @@ -65,6 +65,7 @@
>  #define ARCH_mt
>  #define ARCH_msp430
>  #define ARCH_ns32k
> +#define ARCH_mycpu
>  #define ARCH_openrisc
>  #define ARCH_or32
>  #define ARCH_pdp11
> @@ -319,6 +320,11 @@
>        disassemble = print_insn_openrisc;
>        break;
>  #endif
> +#ifdef ARCH_mycpu
> +    case bfd_arch_mycpu:
> +      disassemble = print_insn_mycpu;
> +      break;
> +#endif
>  #ifdef ARCH_or32
>      case bfd_arch_or32:
>        if (bfd_big_endian (abfd))
>