thanks£¬great work !
make an all out effort !

dolive Wrote:

> thanks£¬great work !
> make an all out effort !

I've implemented 127 missing methods (958 down to 831) yesterday.

druntime compiles for a long time now (producing exactly the same binaries that DMD produces) and I hope that Phobos will be there soon enough.

I'd say it is 80% complete but still missing a few key components.
I have only tested it on Win32, but Linux might work, too (the was some work done to support it).

Keep in mind that it is based on DMD2.032 which is almost a year old. It is also D2-only ATM but I believe D1 support can be added with little effort since a lot of the code is shared between the two versions with differences being versioned out. I tried not to break anything as much as I could.

I'll report more when I manage to compile Phobos entirely.

On 21/08/10 21:32, Denis Koroskin wrote:
> I'd say it is 80% complete but still missing a few key components.
> I have only tested it on Win32, but Linux might work, too (the was some work done to support it).

I got it compiling on linux, and working in a lot of the cases dmd does, I was hitting some random SIGSEGVs in places which I couldn't track down though, I don't remember the specifics, and I didn't spend much time on it - if someone wanted to pick up where I left off they could probably track it down/fix it fairly quickly if familiar with the dmd source code.

You should also note that the instructions for building on linux are slightly wrong in that you need to comment out the dmd main() function, before compiling the backend library, otherwise you end up using dmd rather than ddmd.

-- 
Robert
http://octarineparrot.com/

== Quote from Denis Koroskin (2korden@gmail.com)'s article
> dolive Wrote:
> > thanks£¬great work !
> > make an all out effort !
> I've implemented 127 missing methods (958 down to 831) yesterday.
> druntime compiles for a long time now (producing exactly the same binaries that
DMD produces) and I hope that Phobos will be there soon enough.
> I'd say it is 80% complete but still missing a few key components.
> I have only tested it on Win32, but Linux might work, too (the was some work
done to support it).
> Keep in mind that it is based on DMD2.032 which is almost a year old. It is also
D2-only ATM but I believe D1 support can be added with little effort since a lot of the code is shared between the two versions with differences being versioned out. I tried not to break anything as much as I could.
> I'll report more when I manage to compile Phobos entirely.

So when ddmd finally works fully, is Walter going to ditch the C++ codebase and
base the official DMD implementation on ddmd?  This would be great for a few reasons:

1.  The PR benefits of having a self-hosting D compiler.

2.  If Walter is eating his own dogfood, it will give him a better feel for what bugs are high vs. low priority.

3.  If the productivity of D is higher than C++ by as much as we think it is, DMD should progress faster after the switch is made.  I have my doubts, though, about the productivity benefits of D when you throw legacy code translated from old-school C++ into the mix.

4.  I probably speak for a lot of people on this one, and one reason why I don't contribute to DMD even though I contribute to Phobos is because DMD is in C++ and I don't know C++ thoroughly.  I know the basic syntax and semantics of the language, but I'm not so familiar with the idioms, conventions, gotchas and corner cases.  I'd be a lot more inclined to help out with DMD if it were written in D.

"dsimcha" <dsimcha@yahoo.com> wrote in message news:i4pf6d$15e1$1@digitalmars.com...
> == Quote from Denis Koroskin (2korden@gmail.com)'s article
>> dolive Wrote:
>> > thanksgreat work !
>> > make an all out effort !
>> I've implemented 127 missing methods (958 down to 831) yesterday.
>> druntime compiles for a long time now (producing exactly the same
>> binaries that
> DMD produces) and I hope that Phobos will be there soon enough.
>> I'd say it is 80% complete but still missing a few key components.
>> I have only tested it on Win32, but Linux might work, too (the was some
>> work
> done to support it).
>> Keep in mind that it is based on DMD2.032 which is almost a year old. It is also
> D2-only ATM but I believe D1 support can be added with little effort since
> a lot
> of the code is shared between the two versions with differences being
> versioned
> out. I tried not to break anything as much as I could.
>> I'll report more when I manage to compile Phobos entirely.
>
> So when ddmd finally works fully, is Walter going to ditch the C++
> codebase and
> base the official DMD implementation on ddmd?  This would be great for a
> few reasons:
>
> 1.  The PR benefits of having a self-hosting D compiler.
>
> 2.  If Walter is eating his own dogfood, it will give him a better feel
> for what
> bugs are high vs. low priority.
>
> 3.  If the productivity of D is higher than C++ by as much as we think it
> is, DMD
> should progress faster after the switch is made.  I have my doubts,
> though, about
> the productivity benefits of D when you throw legacy code translated from
> old-school C++ into the mix.
>
> 4.  I probably speak for a lot of people on this one, and one reason why I
> don't
> contribute to DMD even though I contribute to Phobos is because DMD is in
> C++ and
> I don't know C++ thoroughly.  I know the basic syntax and semantics of the
> language, but I'm not so familiar with the idioms, conventions, gotchas
> and corner
> cases.  I'd be a lot more inclined to help out with DMD if it were written
> in D.

All good points, but one problem with moving DMD's official source from C++ to D is it would make it difficult to port DMD to new platforms. Nearly every platform under then sun has a C++ compiler, but not so much for D compilers. Once a good, reliable, well-maintained, up-to-date C- or C++-producing backend for D is available (ideally for DMD itself, as I don't think Walter would want DMD's portability to be reliant on a third-party compiler like LDC or GDC), then that would be the time for DMD to make the switch.

On 8/21/2010 15:18, Nick Sabalausky wrote:
> All good points, but one problem with moving DMD's official source from C++ to D is it would make it difficult to port DMD to new platforms. Nearly every platform under then sun has a C++ compiler, but not so much for D compilers.

Why would that matter?  Why would you want to compile the compiler for a platform, on that platform?


-- 
Rainer Deyke - rainerd@eldwood.com

"Rainer Deyke" <rainerd@eldwood.com> wrote in message news:i4pju0$1pjd$1@digitalmars.com...
> On 8/21/2010 15:18, Nick Sabalausky wrote:
>> All good points, but one problem with moving DMD's official source from
>> C++
>> to D is it would make it difficult to port DMD to new platforms. Nearly
>> every platform under then sun has a C++ compiler, but not so much for D
>> compilers.
>
> Why would that matter?  Why would you want to compile the compiler for a platform, on that platform?
>

To use that platform to compile something.

== Quote from Rainer Deyke (rainerd@eldwood.com)'s article
> On 8/21/2010 15:18, Nick Sabalausky wrote:
> > All good points, but one problem with moving DMD's official source from C++ to D is it would make it difficult to port DMD to new platforms. Nearly every platform under then sun has a C++ compiler, but not so much for D compilers.
> Why would that matter?  Why would you want to compile the compiler for a platform, on that platform?

I have to compile DMD on some computers that I have SSH access to because the sysadmin insists on running some caveman era distro of Linux.  It uses kernel version 2.6.9, which was apparently released in 2004!  The binary that ships with DMD doesn't work with the shared libraries on a distro this old.

Does ddmd use/port the back end? If so, you may be violating the backend license. If not, what are you using as the backend?

Denis Koroskin Wrote:

> dolive Wrote:
> 
> > thanks£¬great work !
> > make an all out effort !
> 
> I've implemented 127 missing methods (958 down to 831) yesterday.
> 
> druntime compiles for a long time now (producing exactly the same binaries that DMD produces) and I hope that Phobos will be there soon enough.
> 
> I'd say it is 80% complete but still missing a few key components.
> I have only tested it on Win32, but Linux might work, too (the was some work done to support it).
> 
> Keep in mind that it is based on DMD2.032 which is almost a year old. It is also D2-only ATM but I believe D1 support can be added with little effort since a lot of the code is shared between the two versions with differences being versioned out. I tried not to break anything as much as I could.
> 
> I'll report more when I manage to compile Phobos entirely.

On 8/21/2010 17:14, Nick Sabalausky wrote:
> "Rainer Deyke" <rainerd@eldwood.com> wrote in message news:i4pju0$1pjd$1@digitalmars.com...
>> Why would that matter?  Why would you want to compile the compiler for a platform, on that platform?
>>
> 
> To use that platform to compile something.

It took me a while to make any sense of that statement.  You're talking about a situation where no compiler binary exists that runs on your platform, so you want to compile the compiler for that platform on the platform itself, right?

The platform the compiler targets and the platform on which the compiler runs are orthogonal issues.  If you want the compiler to produce native binaries on platform X, it must first be able to target platform X.  If the compiler can target platform X, then it is easy enough to create a binary of the compiler that runs on platform X by compiling on platform A.  From there it's a simple matter to write a script that compiles, packages, and uploads the compiler for all supported platforms.  It should therefore never be necessary to compile the compiler itself on platform X.  Or am I missing something?


-- 
Rainer Deyke - rainerd@eldwood.com