Compiler bug? regression for the template function call syntax

2013-05-31 Thread rmc 

Hi all,

Not sure if this is a bug so thought I'd post here.

It is very common to use a function template without the call 
"()" expecting that it will be called. However with the latest 
compilers hibernated only compiles when one leaves out the 
-property dmd command line option.


The call:
getColumnName!(MemberTest, "simple")

Has to be changes to:
getColumnName!(MemberTest, "simple")()

Now do that everywhere a template function is used.

Any help?

Thanks.
Rory

Re: Compiler bug? regression for the template function call syntax

2013-05-31 Thread Ali Çehreli 

On 05/31/2013 07:32 AM, rmc wrote:

> It is very common to use a function template without the call "()"
> expecting that it will be called.

Non-templated functions too.

> However with the latest compilers
> hibernated only compiles when one leaves out the -property dmd command
> line option.

I think it is the same issue with non-templated functions. The best 
thing to do is to forget about the -property switch. I did about a year 
ago and never looked back. :)


If you need to use the -property switch, then define the function with 
the @property attribute:


@property
void getColumnName(T, string s)() {
// ..
}

Ali

Re: Compiler bug? regression for the template function call syntax

2013-05-31 Thread Jonathan M Davis 
On Friday, May 31, 2013 08:27:19 Ali Çehreli wrote:
> I think it is the same issue with non-templated functions. The best
> thing to do is to forget about the -property switch. I did about a year
> ago and never looked back. :)

Given the most recent discussions on @property, I think that it's a foregone 
conclusion that we will never have strict property enforcement (which is what 
-property is trying to do), and -property will be going away at some point in 
the future. The general consensus is that folks want optional parens and not 
have that have anything to do with @property. I think that the only real 
question at this point is what happens with setters, as some folks want 
setters to be restricted to @property, and others just want to go to the 
behavior we had before @property and not need @property at all. For the most 
part, at this point, @property is looking like a failure.

- Jonathan M Davis

Re: Compiler bug? regression for the template function call syntax

2013-05-31 Thread bearophile 

Ali Çehreli:


The best thing to do is to forget about the -property
switch. I did about a year ago and never looked back. :)


But the whole issue of -property was a partial failure for the D 
community and D designers. Several threads, hundreds of posts, 
and I see no clear directives or solution. And the current 
solution is not good.


Bye,
bearophile