Me again:

<code>
int main(){
	return foo(1);
}

struct a {int b;}

int foo(int i) {
	a bar()
	// uncomment for error
	/*out (result) {
		// out body
	}
	body */{
		a r;
		r.b = 1;
		return r;
	}
	
	return bar().b;
}
</code>

you probably need to specify -debug to get it.
Any derived type (structures, arrays, functions
(can you return a function?)) as a return type for
a nested function with any in/out/body block (even just body)
causes this.

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

I don't know what in specific you mean, but there are several syntactical errors here.

"Buchan" <kbuchan@xtra.co.nz> wrote in message news:opsg68fuvz44buww@simon.mshome.net...
> Me again:
>
> <code>
> int main(){
> return foo(1);
> }
>
> struct a {int b;}
>
> int foo(int i) {
> a bar()
> // uncomment for error
> /*out (result) {
> // out body
> }
> body */{
> a r;
> r.b = 1;
> return r;
> }
>
> return bar().b;
> }
> </code>
>
> you probably need to specify -debug to get it.
> Any derived type (structures, arrays, functions
> (can you return a function?)) as a return type for
> a nested function with any in/out/body block (even just body)
> causes this.
>
> -- 
> Using Opera's revolutionary e-mail client: http://www.opera.com/m2/

On Thu, 11 Nov 2004 01:12:01 -0800, Walter <newshound@digitalmars.com> wrote:

> I don't know what in specific you mean, but there are several syntactical
> errors here.
>
> "Buchan" <kbuchan@xtra.co.nz> wrote in message
> news:opsg68fuvz44buww@simon.mshome.net...
>> Me again:
>>
>> <code>
>> int main(){
>> return foo(1);
>> }
>>
>> struct a {int b;}
>>
>> int foo(int i) {
>> a bar()
>> // uncomment for error
>> /*out (result) {
>> // out body
>> }
>> body */{
>> a r;
>> r.b = 1;
>> return r;
>> }
>>
>> return bar().b;
>> }
>> </code>
>> <snip>

Specificly? It compiles fine with it commented, and complains when uncommented.
Is there any reason that a nested function that returns a non-basic type cannot
have in/out contracts? Or am I being really stupid again?

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

"Buchan" <kbuchan@xtra.co.nz> wrote in message news:opshaqkvgr44buww@simon.mshome.net...
> On Thu, 11 Nov 2004 01:12:01 -0800, Walter <newshound@digitalmars.com> wrote:
>
> > I don't know what in specific you mean, but there are several
syntactical
> > errors here.
> >
> > "Buchan" <kbuchan@xtra.co.nz> wrote in message news:opsg68fuvz44buww@simon.mshome.net...
> >> Me again:
> >>
> >> <code>
> >> int main(){
> >> return foo(1);
> >> }
> >>
> >> struct a {int b;}
> >>
> >> int foo(int i) {
> >> a bar()
> >> // uncomment for error

You need an in{} preceding the out{}.

> >> /*out (result) {
> >> // out body
> >> }
> >> body */{
> >> a r;
> >> r.b = 1;
> >> return r;
> >> }
> >>
> >> return bar().b;
> >> }
> >> </code>
> >> <snip>
>
> Specificly? It compiles fine with it commented, and complains when
> uncommented.
> Is there any reason that a nested function that returns a non-basic type
> cannot
> have in/out contracts? Or am I being really stupid again?
>
> -- 
> Using Opera's revolutionary e-mail client: http://www.opera.com/m2/

Walter schrieb am Donnerstag, 11. November 2004 19:53:
> You need an in{} preceding the out{}.

Nested functions require a full set of in&out&body whereas "normal" functions require one of the combinations: body, in&body, out&body or in&out&body.

Is this a bug or a language feature that I missed while reading the documentation?

In addition I might have missed the sequence requirements for in/out/body.

e.g. out/in/body compiles, but in/body/out fails

Thomas

I'll review it again and straighten it out.

"Thomas Kühne" <thomas-dloop@kuehne.cn> wrote in message news:cn0llv$10u5$1@digitaldaemon.com...
> Walter schrieb am Donnerstag, 11. November 2004 19:53:
> > You need an in{} preceding the out{}.
>
> Nested functions require a full set of in&out&body whereas "normal" functions require one of the combinations: body, in&body, out&body or in&out&body.
>
> Is this a bug or a language feature that I missed while reading the documentation?
>
> In addition I might have missed the sequence requirements for in/out/body.
>
> e.g. out/in/body compiles, but in/body/out fails
>
> Thomas
>

On Thu, 11 Nov 2004 10:53:48 -0800, Walter <newshound@digitalmars.com> wrote:

>
> "Buchan" <kbuchan@xtra.co.nz> wrote in message
> news:opshaqkvgr44buww@simon.mshome.net...
>> On Thu, 11 Nov 2004 01:12:01 -0800, Walter <newshound@digitalmars.com>
>> wrote:
>>
<snip>
>
> You need an in{} preceding the out{}.
>
<snip>

Oh? I didnt see that in the documentation... but it does the same even with both.

int main(){
	return foo();
}

struct a {int b;}

int foo() {
	a bar()
/* 	in
	{
		assert(1);
	}
	out (result)
	{
		assert(result.b);
	}
	body*/
	{
		a r;
		r.b = 1;
		return r;
	}

	a r2 = bar();
	return r2.b;
}

If you rewrite this to return a basic type like int, it's fine.
(Of course, I've made an idiot of myself before... ;)

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

added to DStress as
http://svn.kuehne.cn/dstress/run/in_out_body_01.d
...
http://svn.kuehne.cn/dstress/run/in_out_body_04.d

Thomas

Buchan schrieb am Sonntag, 14. November 2004 07:50:
> Oh? I didnt see that in the documentation... but it does the same even with both.
> 
> int main(){
> return foo();
> }
> 
> struct a {int b;}
> 
> int foo() {
> a bar()
> /*    in
> {
> assert(1);
> }
> out (result)
> {
> assert(result.b);
> }
> body*/
> {
> a r;
> r.b = 1;
> return r;
> }
> 
> a r2 = bar();
> return r2.b;
> }
> 
> If you rewrite this to return a basic type like int, it's fine. (Of course, I've made an idiot of myself before... ;)

Looking at the code, I see I accidently left my old e-mail in my account...

BTW, I really should use better names for my variables, nice clean (I think
I wrote that after reading some of the DMD frontend code though ]:D)

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