I'm wondering how to work around a linking issue involving templates and separate compilation. I have three files temp1.d temp2.d and temp3.d:

module temp1;
struct Foo(T) {
  void set(T x) { y = x; }
  T y;
}

module temp2;
import temp1;
void t2() {
  Foo!(int)* x;
  x = new Foo!(int);
  x.set(10);
}

module temp3;
import temp1;
import temp2;
void t3() {
  Foo!(int)* x;
  x = new Foo!(int);
  x.set(100);
}
int main() {
  t2();
  t3();
  return 0;
}

Now if I run
 dmd temp1.d temp2.d temp3.d
then everything works fine. But if I run
 dmd -c temp1;d
 dmd -c temp2.d
 dmd -c temp3.d
 dmd temp1.obj temp2.obj temp3.obj
I get:
C:\d>dmd temp1.obj temp2.obj temp3.obj
C:\dmd\bin\..\..\dm\bin\link.exe temp1+temp2+temp3,,,user32+kernel32/noi;
OPTLINK (R) for Win32  Release 7.50B1
Copyright (C) Digital Mars 1989 - 2001  All Rights Reserved

temp3.obj(temp3)  Offset 001C7H Record Type 00C3
 Error 1: Previous Definition Different : _D5temp15Foo_i3Foo3setFiZv
--- errorlevel 1

Is this going to be a fact of life involving templates? Or is this a temporary state? I can redo my MinTL makefiles to always list everything on one command line when building unittests. I'm not sure if it will come up elsewhre. I'll have to check that none of my libraries contain template instantiations that could end up conflicting with user code. It basically makes separate compilation incompatible with templates unless you are very careful to never mix object files that could share a template instantiation.

-Ben

This shouldn't happen.  I would consider it a bug.


Sean

Added to DStress: http://svn.kuehne.cn/dstress/complex/linking/

all 3 in one step:
make complex/linking/linking_01

3 in 3 steps:
make complex/linking/linking_02

Thomas

Ben Hinkle schrieb am Mon, 6 Dec 2004 20:04:51 -0500:
> I'm wondering how to work around a linking issue involving templates and separate compilation. I have three files temp1.d temp2.d and temp3.d:
>
> module temp1;
> struct Foo(T) {
>   void set(T x) { y = x; }
>   T y;
> }
>
> module temp2;
> import temp1;
> void t2() {
>   Foo!(int)* x;
>   x = new Foo!(int);
>   x.set(10);
> }
>
> module temp3;
> import temp1;
> import temp2;
> void t3() {
>   Foo!(int)* x;
>   x = new Foo!(int);
>   x.set(100);
> }
> int main() {
>   t2();
>   t3();
>   return 0;
> }
>
> Now if I run
>  dmd temp1.d temp2.d temp3.d
> then everything works fine. But if I run
>  dmd -c temp1;d
>  dmd -c temp2.d
>  dmd -c temp3.d
>  dmd temp1.obj temp2.obj temp3.obj
> I get:
> C:\d>dmd temp1.obj temp2.obj temp3.obj
> C:\dmd\bin\..\..\dm\bin\link.exe temp1+temp2+temp3,,,user32+kernel32/noi;
> OPTLINK (R) for Win32  Release 7.50B1
> Copyright (C) Digital Mars 1989 - 2001  All Rights Reserved
>
> temp3.obj(temp3)  Offset 001C7H Record Type 00C3
>  Error 1: Previous Definition Different : _D5temp15Foo_i3Foo3setFiZv
> --- errorlevel 1
>
> Is this going to be a fact of life involving templates? Or is this a temporary state? I can redo my MinTL makefiles to always list everything on one command line when building unittests. I'm not sure if it will come up elsewhre. I'll have to check that none of my libraries contain template instantiations that could end up conflicting with user code. It basically makes separate compilation incompatible with templates unless you are very careful to never mix object files that could share a template instantiation.

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

>> Now if I run
>>  dmd temp1.d temp2.d temp3.d
>> then everything works fine. But if I run
>>  dmd -c temp1;d
>>  dmd -c temp2.d
>>  dmd -c temp3.d
>>  dmd temp1.obj temp2.obj temp3.obj
>> I get:
>> C:\d>dmd temp1.obj temp2.obj temp3.obj
>> C:\dmd\bin\..\..\dm\bin\link.exe temp1+temp2+temp3,,,user32+kernel32/noi;
>> OPTLINK (R) for Win32  Release 7.50B1
>> Copyright (C) Digital Mars 1989 - 2001  All Rights Reserved
>>
>> temp3.obj(temp3)  Offset 001C7H Record Type 00C3
>>  Error 1: Previous Definition Different : _D5temp15Foo_i3Foo3setFiZv
>> --- errorlevel 1

Just a word of warning:
GDC successfully compiles them separately, but fails on the combined invocation..

Thomas


-----BEGIN PGP SIGNATURE-----

iD8DBQFBvMsk3w+/yD4P9tIRAs9FAKCGr08EE3ycewH4Vv+nef14zuRheQCgvY9W
C2MFueVLNstEvqenobAg84A=
=vfLw
-----END PGP SIGNATURE-----