GDC is now on SourceForge: http://dgcc.sourceforge.net/.  The source for the new release can be downloaded now.  I will be adding binaries and a subversion repository.

Download here:
http://sourceforge.net/project/showfiles.php?group_id=154306

* Fixes
   o ICE on assignment arithmetic
   o Problems mixing D and assembler floating point operations
   o Mac OS X 10.2 build problems
   o The gdc driver no longer compiles C source as C++
   o Many others...
* Improvements
   o Updated to DMD 0.157
   o Support for PowerPC Linux
   o Partial support for SkyOS
   o Compiler can be relocated to another directory.

If you have sent e-mail and I have not responded yet, please try again.

David

David Friedman escribió:
> GDC is now on SourceForge: http://dgcc.sourceforge.net/.  The source for the new release can be downloaded now.  I will be adding binaries and a subversion repository.
> 
> Download here:
> http://sourceforge.net/project/showfiles.php?group_id=154306
> 
> * Fixes
>    o ICE on assignment arithmetic
>    o Problems mixing D and assembler floating point operations
>    o Mac OS X 10.2 build problems
>    o The gdc driver no longer compiles C source as C++
>    o Many others...
> * Improvements
>    o Updated to DMD 0.157
>    o Support for PowerPC Linux
>    o Partial support for SkyOS
>    o Compiler can be relocated to another directory.
> 
> If you have sent e-mail and I have not responded yet, please try again.
> 
> David

Great!

"gdc -v" says it's using DMD 0.149, but it does seem to be 0.157.
Also, I couldn't compile Mango on Linux for PowerPC, but I have to try to get a minimal case (unless someone does it first).

Thanks!

-- 
Carlos Santander Bernal

David Friedman wrote:
> GDC is now on SourceForge: http://dgcc.sourceforge.net/.  The source for the new release can be downloaded now.  I will be adding binaries and a subversion repository.
> 
> Download here:
> http://sourceforge.net/project/showfiles.php?group_id=154306
> 
> * Fixes
>    o ICE on assignment arithmetic
>    o Problems mixing D and assembler floating point operations
>    o Mac OS X 10.2 build problems
>    o The gdc driver no longer compiles C source as C++
>    o Many others...
> * Improvements
>    o Updated to DMD 0.157
>    o Support for PowerPC Linux
>    o Partial support for SkyOS
>    o Compiler can be relocated to another directory.
> 
> If you have sent e-mail and I have not responded yet, please try again.
> 
> David

Thanks, David!

-JJR

Good to hear that GDC is absolutely active. Thanks.

Is it technically possible to build a gdc in D?
E.g with the sources of the dparser project?

Frank Benoit wrote:
> Good to hear that GDC is absolutely active. Thanks.
> 
> Is it technically possible to build a gdc in D?
> E.g with the sources of the dparser project?

It is possible.  I think interfacing the whole compiler directly to D would not be trivial, but writing most of the compiler in D (semantic processing) and a smaller part in C (generating GCC tree nodes) should be workable.

David

On Sat, 27 May 2006, David Friedman wrote:

> Frank Benoit wrote:
> > Good to hear that GDC is absolutely active. Thanks.
> > 
> > Is it technically possible to build a gdc in D?
> > E.g with the sources of the dparser project?
> 
> It is possible.  I think interfacing the whole compiler directly to D would not be trivial, but writing most of the compiler in D (semantic processing) and a smaller part in C (generating GCC tree nodes) should be workable.
> 
> David

Though, that means an even longer, harder, upgrade path as dmd is upgraded.  Right now, gdc uses dmd almost directly, and with some time and effort, merging gdc's patches back to dmd will make that even easier.  By having to go and update a d implementation of dmd before gdc can use it, the work just gets harder with little gain.  Now, if DMD were to convert to D, thus using itself to build itself, then I can see the benefits.

I can see some benefits to moving the implementation to D, but there's an awful lot of costs that seem to me to greatly how way the benefits.

Do you see benefits that justify the costs that I'm missing?

Later,
Brad

David Friedman wrote:
> GDC is now on SourceForge: http://dgcc.sourceforge.net/.  The source for the new release can be downloaded now.  I will be adding binaries and a subversion repository.
> 
> Download here:
> http://sourceforge.net/project/showfiles.php?group_id=154306
> 
> * Fixes
>    o ICE on assignment arithmetic
>    o Problems mixing D and assembler floating point operations
>    o Mac OS X 10.2 build problems
>    o The gdc driver no longer compiles C source as C++
>    o Many others...
> * Improvements
>    o Updated to DMD 0.157
>    o Support for PowerPC Linux
>    o Partial support for SkyOS
>    o Compiler can be relocated to another directory.
> 
> If you have sent e-mail and I have not responded yet, please try again.
> 
> David

Awesome, Thanks!

I'll be merging all of my WinCE stuff into this if all goes well.  I have managed to get gphobos to compile for arm-wince gdc 0.17, but there are linking problems at the moment.

> I can see some benefits to moving the implementation to D, but there's an awful lot of costs that seem to me to greatly how way the benefits.
> 
> Do you see benefits that justify the costs that I'm missing?


I thought about that the code would be more attractive to contributors.
Everybody who has interest in gdc is interested in coding in D. So it
would be nice to have gdc in D.
New technics can be tested, using D and D libraries.
And dparser would always be up to date and verified in practice.
I am sure, the time will come and dmd source will switch to D.

Frank Benoit schrieb:
>> I can see some benefits to moving the implementation to D, but there's an awful lot of costs that seem to me to greatly how way the benefits.
>>
>> Do you see benefits that justify the costs that I'm missing?
> 
> 
> I thought about that the code would be more attractive to contributors.
> Everybody who has interest in gdc is interested in coding in D. So it
> would be nice to have gdc in D.
> New technics can be tested, using D and D libraries.
> And dparser would always be up to date and verified in practice.
> I am sure, the time will come and dmd source will switch to D.

Maybe one could have it in parallel - so you could be experimenting, but of course you'd have to do most of the bugfixes two times. But maybe it's better to start it off as a parallel, experimental development, still having the "direct" dmd interface as default.

Mr. Friedman,
Your work is _GREATLY_ apriciated throughout whole D community. I can only write THANKS YOU, THANK YOU, THANK YOU for your work!

Kind regards

Dejan Lekic