On Mon, Feb 16, 2009 at 7:27 PM, Jarrett Billingsley <jarrett.billingsley@gmail.com> wrote:

>

So byte[] (Chris Miller, sans R.) suggested that I suggest that we get a flag for the Windows version of DMD to output ELF objects.  The ELF writer is already there.  Why not just give us the option, and let us fend for ourselves?  ;)

Jarrett Billingsley wrote:
> So byte[] (Chris Miller, sans R.) suggested that I suggest that we get
> a flag for the Windows version of DMD to output ELF objects.  The ELF
> writer is already there.  Why not just give us the option, and let us
> fend for ourselves?  ;)

There's more to it than that, such as the memory model. For example, are reals 10 bytes (Windows) or 12 (Linux) or 16 (OSX) ?

On Thu, Feb 19, 2009 at 10:44 PM, Walter Bright <newshound1@digitalmars.com> wrote:
> Jarrett Billingsley wrote:
>>
>> So byte[] (Chris Miller, sans R.) suggested that I suggest that we get a flag for the Windows version of DMD to output ELF objects.  The ELF writer is already there.  Why not just give us the option, and let us fend for ourselves?  ;)
>
> There's more to it than that, such as the memory model. For example, are
> reals 10 bytes (Windows) or 12 (Linux) or 16 (OSX) ?
>

So I'm assuming you have code like:

#if USING_ELF
    real.alignof = 12;
#elseif USING_OMF
    real.alignof = 10;
#elseif USING_MACH_O
    real.alignof = 16;
#endif

So.. all you'd have to do is separate the "OS" logical decisions from the "object type" decisions.

Or?

Jarrett Billingsley wrote:
> So.. all you'd have to do is separate the "OS" logical decisions from
> the "object type" decisions.
> 
> Or?

Let me put it this way - I thought doing the mac port, which is stuff like that, would take a week. It took 6 weeks.

An elf output would be targeting a different C/C++ compiler ABI and runtime library. It's not terribly difficult, it's just time consuming and not trivial.

On Thu, Feb 19, 2009 at 11:47 PM, Walter Bright <newshound1@digitalmars.com> wrote:
> Jarrett Billingsley wrote:
>>
>> So.. all you'd have to do is separate the "OS" logical decisions from the "object type" decisions.
>>
>> Or?
>
> Let me put it this way - I thought doing the mac port, which is stuff like that, would take a week. It took 6 weeks.
>
> An elf output would be targeting a different C/C++ compiler ABI and runtime library. It's not terribly difficult, it's just time consuming and not trivial.
>

But at the same time, the benefits are undeniable.  s/ELF/any object format other than OMF/, and it still holds.

Walter Bright Wrote:

> Let me put it this way - I thought doing the mac port, which is stuff like that, would take a week. It took 6 weeks.

... so we can expect a 64 bit dmd next week? <g>

在 Fri, 20 Feb 2009 12:59:17 +0800，Jarrett Billingsley <jarrett.billingsley@gmail.com> 写道:

> On Thu, Feb 19, 2009 at 11:47 PM, Walter Bright
> <newshound1@digitalmars.com> wrote:
>> Jarrett Billingsley wrote:
>>>
>>> So.. all you'd have to do is separate the "OS" logical decisions from
>>> the "object type" decisions.
>>>
>>> Or?
>>
>> Let me put it this way - I thought doing the mac port, which is stuff like
>> that, would take a week. It took 6 weeks.
>>
>> An elf output would be targeting a different C/C++ compiler ABI and runtime
>> library. It's not terribly difficult, it's just time consuming and not
>> trivial.
>>
>
> But at the same time, the benefits are undeniable.  s/ELF/any object
> format other than OMF/, and it still holds.

I've got an idea is that DDL linker load everything, then a PE dump(a lot debugger plugin can do this) to create the binary with some little post patch.