• Daniel Green (26/26) Dec 05 2011 Please post all issues in D.gnu or on GDC's bitbucket site
• Trass3r (7/15) Dec 05 2011 Imho the 64 bit version is even more important since gdc is the only
• Daniel Green (6/8) Dec 05 2011 Because this is the first release where I felt D2 was capable of being
• Trass3r (2/9) Dec 05 2011 I see.
• =?UTF-8?B?QWxleCBSw7hubmUgUGV0ZXJzZW4=?= (4/14) Dec 06 2011 That would be very helpful, indeed! I don't feel like rewriting all of
• bearophile (17/18) Dec 05 2011 I suggest to make D2 the default.
• Daniel Green (7/21) Dec 05 2011 Debug information. That's where it was compiled.
• Daniel Green (7/7) Dec 10 2011 Updated binaries.
• bearophile (7/9) Dec 10 2011 I have tried the same install operations I have tried in a precedent pos...
• Daniel Green (7/9) Dec 10 2011 Did you install TDM-GCC or MinGW Stable?
• bearophile (14/20) Dec 10 2011 Like in my precedent test I have installed "tdm-gcc-4.6.1 exe MinGW / sj...
• Daniel Green (7/9) Dec 11 2011 Sorry, it's really a packaging issue. Switching GDC to target i686
• Andrew Wiley (8/19) Dec 11 2011 o
• bearophile (11/14) Dec 11 2011 Now it compiles a working hello world, thank you :-)
• Daniel Green (6/13) Dec 11 2011 I recall MinGW may include extra object files during the link process.
• Andrej Mitrovic (3/3) Dec 12 2011 This is great, the binaries now work. Thanks for your hard work Daniel.
• Andrej Mitrovic (3/3) Dec 12 2011 Here's what I can add on line 434 in GDMD to make it work:

Daniel Green <venix1 gmail.com> writes:

Please post all issues in D.gnu or on GDC's bitbucket site 
https://bitbucket.org/goshawk/gdc

  * Built against TDM GCC 4.6.1:  http://tdm-gcc.tdragon.net/
  * Built against GDC revision 7e22befef29c.
  * Working TLS support.
  * Updated binutils and mingw runtime for TLS support.
    * new gas directive  secrel32 will generate a constant offset usable
      as an immediate or displacement value.
      Works like posix counterparts  tpoff,  dtpoff, etc.
    * Fixes an error with the mingw runtime related to TLS.
  * 7-zip format for size reasons:  http://7-zip.org
  * Release includes both D versions.
    * -v1(default) compiles for D1.
    * -v2 compiles for D2.
    * The switch must be used for linking as well.
  * Made possible by Iain Buclaw's hard work in maintaining GDC.
  * 64bit and up to date GDC forthcoming.

To install simply extract to you TDM MinGW32 directory.

This release
https://bitbucket.org/goshawk/gdc/downloads/gcc-4.6.1-tdm-1-gdc-7e22befef29c-20111205.zip

All MinGW GDC downloads.
It's highly recommended to ignore all prior builds.  TLS *will* not work.
https://bitbucket.org/goshawk/gdc/downloads


All patches, source files and build scripts can be found at 
https://bitbucket.org/venix1/mingw-gdc/

Enjoy.

Trass3r <un known.com> writes:

Thx!

   * Release includes both D versions.
     * -v1(default) compiles for D1.
     * -v2 compiles for D2.
     * The switch must be used for linking as well.

Why is D1 still the default?

   * 64bit and up to date GDC forthcoming.

Imho the 64 bit version is even more important since gdc is the only  
option at all to produce Win64 binaries.
(and of course beginning with Win7 everyone should use an x64 OS anyway ;))

   * 7-zip format for size reasons:  http://7-zip.org

perfect.

 This release
 https://bitbucket.org/goshawk/gdc/downloads/gcc-4.6.1-tdm-1-gdc-7e22befef29c-20111205.zip

but why is there a zip version anyway?

Daniel Green <venix1 gmail.com> writes:

On 12/5/2011 8:28 PM, Trass3r wrote:
 Why is D1 still the default?

Because this is the first release where I felt D2 was capable of being 
the default and I forgot about it until writing the post.  It also 
requires some reworking of the changes that enable dual compilers.

 but why is there a zip version anyway?

I posted with a zip extension in the original post, so rather than 
letting everyone get an 404 error, I just uploaded a zip file.

Trass3r <un known.com> writes:

 Why is D1 still the default?

 Because this is the first release where I felt D2 was capable of being  
 the default and I forgot about it until writing the post.  It also  
 requires some reworking of the changes that enable dual compilers.

 but why is there a zip version anyway?

 I posted with a zip extension in the original post, so rather than  
 letting everyone get an 404 error, I just uploaded a zip file.

I see.
btw, is there something like gdmd for Windoze too?

=?UTF-8?B?QWxleCBSw7hubmUgUGV0ZXJzZW4=?= <xtzgzorex gmail.com> writes:

On 06-12-2011 02:46, Trass3r wrote:
 Why is D1 still the default?

 Because this is the first release where I felt D2 was capable of being
 the default and I forgot about it until writing the post. It also
 requires some reworking of the changes that enable dual compilers.

 but why is there a zip version anyway?

 I posted with a zip extension in the original post, so rather than
 letting everyone get an 404 error, I just uploaded a zip file.

 I see.
 btw, is there something like gdmd for Windoze too?

That would be very helpful, indeed! I don't feel like rewriting all of 
my makefiles to use GDC's parameter syntax. ;)

- Alex

bearophile <bearophileHUGS lycos.com> writes:

Daniel Green:

     * -v1(default) compiles for D1.

I suggest to make D2 the default.

I have installed the latest TDM GCC in
C:\MinGW

Then I have unpacked the 7-zip over the directories of the MinGW install. 150+
files got overwritten, some executables too.

I have made sure C:\MinGW\bin and C:\MinGW\lib are in the path. C:\MinGW\lib
contains libgphobos.a too.

If now I compile a little program using:
gdc -v2 test.d

It probably doesn't find Phobos:

c:/mingw/bin/../lib/gcc/mingw32/4.6.1/../../../libgphobos2.a(bitmanip.o): In
function `D3std8bitmanip8BitArray7opIndexMxFkZb':
C:\crossdev\gdc\v2\build\mingw32\libphobos/../../../gcc-4.6.1/libphobos
std/bitmanip.d:410: undefined reference to `__sync_fetch_and_add_4'
...
c:/mingw/bin/../lib/gcc/mingw32/4.6.1/../../../libgphobos2.a(bitmanip.o):C:\crossdev\gdc\v2\build\mingw32\libphobos/../../../gcc-4.6.1/libphobos
std/bitmanip.d:410: more undefined references to `__sync_fetch_and_add_4' follow

I don't even know what C:\crossdev\ is, I don't have that directory.

Do you know how to improve this situation?

Bye and thank you,
bearophile

Daniel Green <venix1 gmail.com> writes:

On 12/5/2011 9:47 PM, bearophile wrote:

 I suggest to make D2 the default.

That's the plan.

 I have installed the latest TDM GCC in
 C:\MinGW

 Then I have unpacked the 7-zip over the directories of the MinGW install. 150+
files got overwritten, some executables too.

 I have made sure C:\MinGW\bin and C:\MinGW\lib are in the path. C:\MinGW\lib
contains libgphobos.a too.

 If now I compile a little program using:
 gdc -v2 test.d

 It probably doesn't find Phobos:

 c:/mingw/bin/../lib/gcc/mingw32/4.6.1/../../../libgphobos2.a(bitmanip.o): In
function `D3std8bitmanip8BitArray7opIndexMxFkZb':
 C:\crossdev\gdc\v2\build\mingw32\libphobos/../../../gcc-4.6.1/libphobos
std/bitmanip.d:410: undefined reference to `__sync_fetch_and_add_4'
 ...
 c:/mingw/bin/../lib/gcc/mingw32/4.6.1/../../../libgphobos2.a(bitmanip.o):C:\crossdev\gdc\v2\build\mingw32\libphobos/../../../gcc-4.6.1/libphobos
std/bitmanip.d:410: more undefined references to `__sync_fetch_and_add_4' follow

 I don't even know what C:\crossdev\ is, I don't have that directory.

Debug information.  That's where it was compiled.

 Do you know how to improve this situation?

I had to recompile binutils to enable TLS support.  It appears I messed 
up some configure options.  I will post once I've fixed the issue and 
uploaded new binaries.

-Daniel

Daniel Green <venix1 gmail.com> writes:

Updated binaries.
https://bitbucket.org/goshawk/gdc/downloads/gcc-4.6.1-tdm-1-gdc-7e22befef29c-20111210.7z

  * GDC now targets i686 processors.  Was targeting i386 which required 
substituting functions for atomic instructions.

I will release MinGW64 binaries soon then work on updating to tip and 
changing the default compiler to D2.

-Daniel

bearophile <bearophileHUGS lycos.com> writes:

Daniel Green:

 Updated binaries.
 https://bitbucket.org/goshawk/gdc/downloads/gcc-4.6.1-tdm-1-gdc-7e22befef29c-20111210.7z

I have tried the same install operations I have tried in a precedent post, and
now it says it is not finding this:
libgmp-3.dll

This thread that seems related:
http://old.nabble.com/GCJ---libgmp-3.dll-was-not-found-td25234066.html

Bye,
bearophile

Daniel Green <venix1 gmail.com> writes:

On 12/10/2011 4:38 PM, bearophile wrote:
 I have tried the same install operations I have tried in a precedent post, and
now it says it is not finding this:
 libgmp-3.dll

Did you install TDM-GCC or MinGW Stable?

TDM-GCC is built with shared libraries using gmp-4.3.2.  MinGW stable is 
built with static libraries.  The dll you listed comes from gmp-5.0.1. 
I am using the TDM-GCC build scripts to build the D compiler as well. 
You can find the files inside the "libexec/mingw32/4.6.1" folder of the 
TDM-GCC core download from the downloads page.

bearophile <bearophileHUGS lycos.com> writes:

Daniel Green:

 Did you install TDM-GCC or MinGW Stable?

Like in my precedent test I have installed "tdm-gcc-4.6.1 exe MinGW / sjlj":
http://sourceforge.net/projects/tdm-gcc/files/TDM-GCC%20Installer/tdm-gcc-4.6.1.exe/download


 TDM-GCC is built with shared libraries using gmp-4.3.2.  MinGW stable is 
 built with static libraries.  The dll you listed comes from gmp-5.0.1. 
 I am using the TDM-GCC build scripts to build the D compiler as well. 
 You can find the files inside the "libexec/mingw32/4.6.1" folder of the 
 TDM-GCC core download from the downloads page.

I have found that libgmp-3.dll and other needed dlls are in this directory,
that I have added to the path:
C:\MinGW32\libexec\gcc\mingw32\4.6.1

Now it's giving me:

...>gdc -v2 test.d
gdc: fatal error: -fuse-linker-plugin, but liblto_plugin-0.dll not found
compilation terminated.

Despite the C:\MinGW32\libexec\gcc\mingw32\4.6.1 directory contains
liblto_plugin-0.dll too.

I feel stupid -.-
I'd like a single zip/rar/7zip with all needed files inside, with GCC too :-)

Bye,
bearophile

Daniel Green <venix1 gmail.com> writes:

On 12/11/2011 2:39 AM, bearophile wrote:
 I feel stupid -.-

Sorry, it's really a packaging issue.  Switching GDC to target i686 
changed some library and exectuable paths resulting in these issues.


 I'd like a single zip/rar/7zip with all needed files inside, with GCC too :-)

That could be done but I never liked the idea of requiring multiple GCC 
installations on any system.  That's why I chose this path.

I've repackaged and tested it from a clean install.

https://bitbucket.org/goshawk/gdc/downloads/gcc-4.6.1-tdm-1-gdc-7e22befef29c-20111211.7z

Andrew Wiley <wiley.andrew.j gmail.com> writes:

On Sun, Dec 11, 2011 at 10:29 AM, Daniel Green <venix1 gmail.com> wrote:
 On 12/11/2011 2:39 AM, bearophile wrote:
 I feel stupid -.-

 Sorry, it's really a packaging issue. =A0Switching GDC to target i686 cha=

nged
 some library and exectuable paths resulting in these issues.



 I'd like a single zip/rar/7zip with all needed files inside, with GCC to=


o
 :-)

 That could be done but I never liked the idea of requiring multiple GCC
 installations on any system. =A0That's why I chose this path.

 I've repackaged and tested it from a clean install.

 https://bitbucket.org/goshawk/gdc/downloads/gcc-4.6.1-tdm-1-gdc-7e22befef=

29c-20111211.7z

This one works for me on TDM GCC. I had the same gmp problem with the last =
one.

It seems calling spawn() in std.concurrency causes a segfault, but
that's a separate issue.

bearophile <bearophileHUGS lycos.com> writes:

Daniel Green:

 I've repackaged and tested it from a clean install.
 
 https://bitbucket.org/goshawk/gdc/downloads/gcc-4.6.1-tdm-1-gdc-7e22befef29c-20111211.7z

Now it compiles a working hello world, thank you :-)

A not stripped hello world is 3_908_177 bytes, much bigger than the binary
produced by DMD, and it's 677_888 bytes stripped.
DMD 2.057beta+optilink produce a 932_892 bytes hello world binary, while using
the same DMD+ulink it becomes 334_576 bytes.

If I compile the little D coins program here:
http://www.digitalmars.com/webnews/newsgroups.php?art_group=digitalmars.D&article_id=151610

Using:
gdc -v2 coins.d -o coins.exe

When I run it it crashes immediately with no error messages :-(

Bye,
bearophile

Daniel Green <venix1 gmail.com> writes:

On 12/11/2011 7:33 PM, bearophile wrote:
 A not stripped hello world is 3_908_177 bytes, much bigger than the binary
produced by DMD, and it's 677_888 bytes stripped.
 DMD 2.057beta+optilink produce a 932_892 bytes hello world binary, while using
the same DMD+ulink it becomes 334_576 bytes.

I recall MinGW may include extra object files during the link process. 
Try comparing against what GCC's C/C++ compiler produces.


 If I compile the little D coins program here:
 http://www.digitalmars.com/webnews/newsgroups.php?art_group=digitalmars.D&article_id=151610

 Using:
 gdc -v2 coins.d -o coins.exe

 When I run it it crashes immediately with no error messages :-(

It's a TLS issue being looked into.  Offsets are being generated against 
the wrong section.

-Dan

Andrej Mitrovic <andrej.mitrovich gmail.com> writes:

This is great, the binaries now work. Thanks for your hard work Daniel.

Btw, it seems GDMD is hardcoded to D1? -fd-version is always 1,
there's a -v1 switch for GDMD but no -v2 switch. Should I file this?

Andrej Mitrovic <andrej.mitrovich gmail.com> writes:

Here's what I can add on line 434 in GDMD to make it work:
    } elsif ( $arg =~ m/^-v2$/ ) {
        push  out, '-v2';