Hi there,

I just ported Gregors patch for DMD 1.018 to the latest SVN.

If any brave soul tries to compile it and succeeds, maybe it can be incorporated.

http://sourceforge.net/tracker/index.php?func=detail&aid=1757177&group_id=154306&atid=791254

I'm going to work on getting further there...

Best Regards

Ingo Oeser, now counting Ds^Wsheeps...

Ingo Oeser wrote:
> Hi there,
> 
> I just ported Gregors patch for DMD 1.018 to the latest SVN.
> 
> If any brave soul tries to compile it and succeeds, maybe it
> can be incorporated.
> 
> http://sourceforge.net/tracker/index.php?func=detail&aid=1757177&group_id=154306&atid=791254
> 
> I'm going to work on getting further there...
> 
> Best Regards
> 
> Ingo Oeser, now counting Ds^Wsheeps...

When I made my 1.018 patch, I actually worked against SVN, I just made the patchfile against 0.23 because it's not a moving target.

 - Gregor Richards

PS: This message may or may not be irrelevant :)

Ingo Oeser wrote:

> If any brave soul tries to compile it and succeeds, maybe it
> can be incorporated.

Hi,

On my powerpc mac with os x.4.10, building with apple's gcc 4.0.1 on the source of apple's gcc sources rev 5363 with gdc rev 127 and this patch, passing --disable-multilib to the configure script.

It fails to compile because in the file d/dmd/inline.c :
int AssocArrayLiteralExp::inlineCost(InlineCostState*),
int StructLiteralExp::inlineCost(InlineCostState*),
Expression* AssocArrayLiteralExp::doInline(InlineDoState*),
Expression* StructLiteralExp::doInline(InlineDoState*),
Expression* AssocArrayLiteralExp::inlineScan(InlineScanState*)
and Expression* StructLiteralExp::inlineScan(InlineScanState*)
were declared twice,
and there was another error about struct Statement not having a member function named "isIfStatement" that should have been used to avoid dynamic_cast, according to the comments.

I removed the duplicate methods and added isIfStatement, then the front-end is ok, but then when compiling libphobos I get :

../../../gcc-5363/libphobos/std/string.d:818: Error: identifier 'string' is not defined

Well, it's late, I'm going to sleep...

Hi Gregor,

Gregor Richards wrote:
> When I made my 1.018 patch, I actually worked against SVN, I just made the patchfile against 0.23 because it's not a moving target.

Yes, but I want more rapid progress in that area. I'm coming
from linux kernel development and that is about 65 commits per day :-)

Until I know the innards of GDC and and DMD better, I just try to help in merging and integrating latest versions of DMD frontent and GCC.

Next step will be dtest nightly run with these versions.

Currently GDC seems to be a one man show, which is not good for productivity even if David does a pretty good job at the moment!

Thanks David!

The GCC 4.2.0 stuff stopped with segfaults. Since GCC is releasing 4.2.1 soon, I will wait for that.

Best regard

Ingo Oeser, insane bleeding edge D junkie :-)

Hi alex,

alex wrote:
> On my powerpc mac with os x.4.10, building with apple's gcc 4.0.1 on the source of apple's gcc sources rev 5363 with gdc rev 127 and this patch, passing --disable-multilib to the configure script.

Ok, I just made some mistakes generating the patch. Sorry for that! So I applied every reject manually, carefully avoiding these goof ups.

I also tested compile with official GCC 4.1.2 repository with Linux/Ubuntu now.

./configure --prefix=/opt/gcc-4.1.2 --disable-multilib --disable-shared --enable-languages=d --disable-nls

I had to fix a reject (due to whitepace?) in Davids part.

Would you mind testing the new version (against SVN r132 of this patch?)

http://sourceforge.net/tracker/index.php?func=detail&aid=1758462&group_id=154306&atid=791254

Many thanks for testing. I really appreciate it, since I cannot test Windows or MacOSX.


Best Regards

Ingo Oeser