Unfortunately, I've found several funky bugs with my patch, so I'm going to do a much more careful incremental patch, using a few real test programs that are known to fail.  See if I can avoid doing that again :)

 - Gregor Richards

Gregor Richards wrote:

> Unfortunately, I've found several funky bugs with my patch, so I'm going to do a much more careful incremental patch, using a few real test programs that are known to fail.  See if I can avoid doing that again :)

The generated gcc.config for Darwin was still broken in it, too...
It doesn't include the "private import std.c.darwin.ldblcompat;" ?

There also needs to be some options added to the driver and wrapper:
http://d.puremagic.com/bugzilla/show_bug.cgi?id=2


But I got std.progress and std.c.fenv ported to Darwin, at least.
The former was just a simple change from "linux" over to "Unix",
the latter was a copy and paste from the the system's C headers.
Posting them as patches to the Bugzilla, once they're cleaned up.

--anders

Gregor Richards wrote:
> Unfortunately, I've found several funky bugs with my patch, so I'm going to do a much more careful incremental patch, using a few real test programs that are known to fail.  See if I can avoid doing that again :)
> 
>  - Gregor Richards

One bug came up in the 147-148 change.  Likely by no coincidence, this is also when the bit->bool change came into effect.

I've posted the working 147 and broken 148 patches to the D Journal:

http://www.tdjonline.com/?page_id=11

 - Gregor Richards

PS: Anders: Once I get back to 156, I'll integrate your patches.  Don't want to make the increments too complicated, sorry :)

Gregor Richards wrote:

> PS: Anders: Once I get back to 156, I'll integrate your patches.
> Don't want to make the increments too complicated, sorry :)

Not a problem, so far they all apply equally well to DMD and GDC.
(there are just three, one to "mars.h" and two to Phobos modules)

I will probably just build and test for Mac, not develop anything.
At least not until Mac OS X 10.5, which is likely to have GCC 4.1...


Beyond the earlier mentioned Linux-PowerPC (and possibly X64 too???)
support, the most needed items for GDC are: DMD 0.156 and GCC 4.1.0

Well, that and the non-technical parts such as finalizing the name
and packaging up for end-users and writing a new home page and docs.

--anders

PS.
Supporting 64-bit might have to wait until DMD does too, I suppose.
Since the new Intel Macs are only 32-bit, it's not *my* big concern :-)

Anders F Björklund wrote:
> Gregor Richards wrote:
> 
>> PS: Anders: Once I get back to 156, I'll integrate your patches.
>> Don't want to make the increments too complicated, sorry :)
> 
> 
> Not a problem, so far they all apply equally well to DMD and GDC.
> (there are just three, one to "mars.h" and two to Phobos modules)
> 
> I will probably just build and test for Mac, not develop anything.
> At least not until Mac OS X 10.5, which is likely to have GCC 4.1...

Nice to know it'll work on Mac at all ;)

> 
> 
> Beyond the earlier mentioned Linux-PowerPC (and possibly X64 too???)
> support, the most needed items for GDC are: DMD 0.156 and GCC 4.1.0

Seems to me like DMD 0.156 is by far the more important issue.  GCC 4.1.0 can wait :)

> 
> Well, that and the non-technical parts such as finalizing the name
> and packaging up for end-users and writing a new home page and docs.

"GCC D Compiler" forever ^^
And hopefully no new homepage will be necessary, since David will reappear and get things back on track.  Optimism is awesome 8-D

> 
> --anders
> 
> PS.
> Supporting 64-bit might have to wait until DMD does too, I suppose.
> Since the new Intel Macs are only 32-bit, it's not *my* big concern :-)

 - Gregor Richards

Gregor Richards wrote:

> Nice to know it'll work on Mac at all ;)

Oh, it does. On the ones using Intel, too...
(at least I think it does, no bug reports yet)

> Seems to me like DMD 0.156 is by far the more important issue.

It is, at least if you want the new things with templates and what not.
You can still run the "traditional" D programs, of course. (DMD 0.140)
But it would be nice if GDC could catch up with the newest D language.

Just meant that it is still useful, even while it haven't done so...

> GCC 4.1.0 can wait :)

The problem being that GCC 4.1.0 is now shipped as the default compiler
on for instance Fedora Core 5, which means that you right now can't even compile GDC on those machines without using a "compat" (GCC 3) compiler.

GCC 4.0 works though, so it should be OK for Mac OS X 10.4 - and Ubuntu.

> And hopefully no new homepage will be necessary, since David will reappear and get things back on track.  Optimism is awesome 8-D

It needs a new* homepage anyway, but I definitely hope David returns!
(* = with the space/bandwidth for the binaries, some design and docs)

We will have to see how it goes, and also hear from Gabe on "gnu-d.org"

--anders

Gregor Richards wrote:
> Likely by no coincidence, this is also when the bit->bool change came into effect.
> 

No coincidence indeed.  I merged only the bit->bool change into the working 147 branch, and the problem persists.  Still investigating.

 - Gregor Richards