Generating headers with -H

[Andrei Alexandrescu]

Steven Schveighoffer wrote:
> On Mon, 02 Nov 2009 11:53:12 -0500, Andrei Alexandrescu 
> <SeeWebsiteForEmail at erdani.org> wrote:
> 
>> This question may actually belong in .learn.
>>
>> What's exactly eliminated from the header generated with -H? I tried 
>> it just now and was surprised to see that the generated .di file 
>> includes the function bodies of regular (non-template) functions and 
>> methods.
>>
>> I was under the impression that generated headers exclude function 
>> bodies, and found documentation to support that viewpoint:
>>
>> http://www.digitalmars.com/d/2.0/dmd-windows.html#interface_files
>>
>> says:
>>
>> ============
>> A D interface file contains only what an import of the module needs, 
>> rather than the whole implementation of that module.
>>
>> The advantages of using a D interface file for imports rather than a D 
>> source file are:
>>
>>      * D interface files are often significantly smaller and much 
>> faster to process than the corresponding D source file.
>>      * They can be used to hide the source code, for example, one can 
>> ship an object code library along with D interface files rather than 
>> the complete source code.
>> ============
>>
>> This strongly suggests that function bodies are eliminated. But what 
>> I'm seeing is that pretty much all function bodies are kept 
>> (exception: static this() functions are not), and only comments are 
>> removed.
>>
>> So what's -H really doing?
> 
> It removes whitespace and comments.
> 
> My understanding of why it does not remove implementations is for CTFE 
> and inlining.
> 
> Somehow I got it in my head that a significantly large enough function 
> would have it's body removed, but I can't remember why I think that...

You're right. I just added a loop to a function, and its body 
disappeared from the .di file. Thanks!

The CTFE-ability of functions seems to play no role in they being kept 
around, however.

> Note that the above snippit from the docs suggests reasons for using d 
> interface files, *NOT* reasons for using dmd -H :)

Yeah, I'd agree if the link didn't come from the same page's explanation 
of -H.

I think the doc should be clearer about the fact that the compiler has 
the freedom to keep whichever function bodies it finds fit for inlining.


Andrei