I've narrowed down some reproduction steps for the "symbol undefined _init_10TypeInfo_i" bug that has been plaguing MinTL lately. If I comment out anything more in the following it compiles ok.

module test1;
template Mix(Key,Value) {
  int opApply1(int delegate(inout Key key, inout Value val) dg) {
    TypeInfo ti_k = typeid(Key);
    TypeInfo ti_v = typeid(Value);
    return 0;
  }
  int opApply2(int delegate(inout Value val) dg) {
    TypeInfo ti_k = typeid(Key);
    return 0;
  }
}
struct F(Key,Value) {
  mixin Mix!(Key,Value);
  alias opApply1 opApply;
  alias opApply2 opApply;
}
int main() {
  F!(int,int) y;
  foreach(int k,int v; y)
    1;
  return 0;
}


C:\d>dmd test1.d
C:\dmd\bin\..\..\dm\bin\link.exe test1,,,user32+kernel32/noi;
OPTLINK (R) for Win32  Release 7.50B1
Copyright (C) Digital Mars 1989 - 2001  All Rights Reserved

test1.obj(test1)
 Error 42: Symbol Undefined _init_10TypeInfo_i
--- errorlevel 1


-Ben

actually the template isn't needed either:

module test1;
struct F(Key,Value) {
  int opApply(int delegate(inout Key key, inout Value val) dg) {
    TypeInfo ti_k = typeid(Key);
    TypeInfo ti_v = typeid(Value);
    return 0;
  }
  int opApply(int delegate(inout Value val) dg) {
    TypeInfo ti_k = typeid(Key);
    //    TypeInfo ti_v = typeid(Value);
    return 0;
  }
}
int main() {
  F!(int,int) y;
  foreach(int k,int v; y)
    1;
  return 0;
}

On Sun, 28 Nov 2004 09:42:18 -0500, Ben Hinkle <ben.hinkle@gmail.com> wrote:

> actually the template isn't needed either:
>
> module test1;
> struct F(Key,Value) {
>   int opApply(int delegate(inout Key key, inout Value val) dg) {
>     TypeInfo ti_k = typeid(Key);
>     TypeInfo ti_v = typeid(Value);
>     return 0;
>   }
>   int opApply(int delegate(inout Value val) dg) {
>     TypeInfo ti_k = typeid(Key);
>     //    TypeInfo ti_v = typeid(Value);
>     return 0;
>   }
> }
> int main() {
>   F!(int,int) y;
>   foreach(int k,int v; y)
>     1;
>   return 0;
> }
>

Hmm, didn't see that class template syntax before... cool, but
confusing. (maybe require template <class> <name>(...) {...} ?)

Whoa... I've been playing around with this a bit, got it down to
<code>
module test1;

void funcA() {
	TypeInfo ti_a1 = typeid(int),
	         ti_a2 = typeid(int);
}
void funcB() {
	TypeInfo ti_a = typeid(uint);
}

int main() {
	return 0;
}
</code>

Some things to note:
- Not to do with overloading, but there must be two functions.
- One function needs at least two TypeInfo's, the other needs at
least one.
- TypeInfo's must be instanced and set. (typeid(TypeA); on its own
doesn't cause this, and TypeInfo ti; doesn't either)
- The TypeInfos must be equal in the two TypeInfo function.

This is one of the most specific bugs I've seen... I'll keep hacking
at it to see if it can be reduced more (I quartered it while typing
this :/)

BTW, a {} empty body is clearer than 1; (At least, to me it is)

-- 
Using Opera's revolutionary e-mail client: http://www.opera.com/m2/

On Mon, 29 Nov 2004 20:13:47 +1300, Simon Buchan <currently@no.where> wrote:

<snip>
>
> Some things to note:
> - Not to do with overloading, but there must be two functions.
> - One function needs at least two TypeInfo's, the other needs at
> least one.
> - TypeInfo's must be instanced and set. (typeid(TypeA); on its own
> doesn't cause this, and TypeInfo ti; doesn't either)
> - The TypeInfos must be equal in the two TypeInfo function.
>
> This is one of the most specific bugs I've seen... I'll keep hacking
> at it to see if it can be reduced more (I quartered it while typing
> this :/)
>
> BTW, a {} empty body is clearer than 1; (At least, to me it is)
>

Well, I got it down to:

module test1;

TypeInfo ti_a1 = typeid(int),
         ti_a2 = typeid(int),
         ti_a3 = typeid(uint);

int main() {
	return 0;
}

The first two must be the same. WTF?

-- 
"Unhappy Microsoft customers have a funny way of becoming Linux, Salesforce.com and Oracle customers." - www.microsoft-watch.com: "The Year in Review: Microsoft Opens Up"

I found and fixed the problem, no need to work on it any more! Thanks.

"Walter" <newshound@digitalmars.com> wrote in message news:cofv3n$1fik$2@digitaldaemon.com...
>I found and fixed the problem, no need to work on it any more! Thanks.
>
>

great! thanks. And here I thought the reproduction steps were complicated. I'm embarassed it took so long to narrow down. :-P

On Mon, 29 Nov 2004 11:56:59 -0800, Walter <newshound@digitalmars.com> wrote:

> I found and fixed the problem, no need to work on it any more! Thanks.
>
>

Awww! :D
Didn't seem much like work...

-- 
"Unhappy Microsoft customers have a funny way of becoming Linux,
Salesforce.com and Oracle customers." - www.microsoft-watch.com:
"The Year in Review: Microsoft Opens Up"