www.digitalmars.com         C & C++   DMDScript  

D.gnu - GDC pre-release posted

reply David Friedman <d3rdclsmail_a_ _t_earthlink_d_._t_net> writes:
Lots of new features, not much documentation.  The full release will be 
made when I sync up with the latest DMD version.

http://home.earthlink.net/~dvdfrdmn/d/

Major Improvements:

* Updated to DMD 0.113
* Phobos is now built as target library (i.e., no need for a separate 
build step.)
* Boehm-gc is no longer used and the Java package is no longer required.
* Inline assembler for x86 (some limitations compared to DMD)
* Included Anders Björklund's patches to enable the use of frameworks on 
Darwin.

David
Apr 11 2005
next sibling parent Dejan Lekic <leka entropy.tmok.com> writes:
PERFECT! I prefer GPL projects and all I can say is _THANKS_! New GDC
version is most welcome...

Dejan

-- 
...........
Dejan Lekic
  http://dejan.lekic.org
Apr 11 2005
prev sibling next sibling parent reply Thomas Kuehne <thomas-dloop kuehne.thisisspam.cn> writes:
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

David Friedman schrieb am Tue, 12 Apr 2005 00:03:40 -0400:

 Lots of new features, not much documentation.  The full release will be 
 made when I sync up with the latest DMD version.

 http://home.earthlink.net/~dvdfrdmn/d/

 Major Improvements:

 * Updated to DMD 0.113
 * Phobos is now built as target library (i.e., no need for a separate 
 build step.)
 * Boehm-gc is no longer used and the Java package is no longer required.
 * Inline assembler for x86 (some limitations compared to DMD)
 * Included Anders Björklund's patches to enable the use of frameworks on 
 Darwin.


Thank's a lot for your great work David.

Somehow Phobos' configure set the target_alias but didn't set the
host_alias and as a consequence make tried to create /gcc 
instead of i686-pc-linux-gnu/gcc.

Thomas 


-----BEGIN PGP SIGNATURE-----

iD8DBQFCW8/Z3w+/yD4P9tIRAkG6AJ9Zerq5xY1jM1NyUQJvSCVfYlbqzQCeJy7a
8JPFx7KHK6UIG6MAIPMkY4Q=
=2Hgj
-----END PGP SIGNATURE-----
Apr 12 2005
parent reply =?ISO-8859-1?Q?Anders_F_Bj=F6rklund?= <afb algonet.se> writes:
Thomas Kuehne wrote:


 Somehow Phobos' configure set the target_alias but didn't set the
 host_alias and as a consequence make tried to create /gcc 
 instead of i686-pc-linux-gnu/gcc.


Thanks Thomas, this got me a little further...
(explicitly setting --host and --build params)

Now it seems to be disliking my paths instead:

i386-redhat-linux-gcc -o internal/gc/gc.o -O2 -g -pipe -march=i386 
-funittest -nostdinc -I ../gcc-3.4.3/gcc/d/phobos -I 
../gcc-3.4.3/gcc/d/phobos/internal/gc -I ./i386-redhat-linux -c 
../gcc-3.4.3/gcc/d/phobos/internal/gc/gc.d
{standard input}: Assembler messages:
{standard input}:10: Error: unknown pseudo-op: `..'
make: *** [internal/gc/gc.o] Error 1

It's really, really hard to try to pass the
standard CFLAGS and DFLAGS to anything in D ?
(having similar problems with DMD and recls)

--anders
Apr 12 2005
parent reply Thomas Kuehne <thomas-dloop kuehne.thisisspam.cn> writes:
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Anders F Björklund schrieb am Tue, 12 Apr 2005 15:56:25 +0200:

 Now it seems to be disliking my paths instead:

 i386-redhat-linux-gcc -o internal/gc/gc.o -O2 -g -pipe -march=i386 
 -funittest -nostdinc -I ../gcc-3.4.3/gcc/d/phobos -I 
 ../gcc-3.4.3/gcc/d/phobos/internal/gc -I ./i386-redhat-linux -c 
 ../gcc-3.4.3/gcc/d/phobos/internal/gc/gc.d
 {standard input}: Assembler messages:
 {standard input}:10: Error: unknown pseudo-op: `..'
 make: *** [internal/gc/gc.o] Error 1


guessing from the error message your directory setup would be

somewhere/gcc-3.4.3
somewhere/gcc-3.4.3/gcc/d/phobos
somewhere/phobos [you are here and trying to build Phobos]

Is that right?

Thomas


-----BEGIN PGP SIGNATURE-----

iD8DBQFCW9j93w+/yD4P9tIRAtdVAKCaxfmIifnbwZTuKdQGSIf4SEXNwACfXCBl
9YM/Vgdo3SSoVv4rT2Q8rRQ=
=Ujhp
-----END PGP SIGNATURE-----
Apr 12 2005
next sibling parent reply David Friedman <d3rdclsmail_a t_earthlink_d.t_net> writes:
Thomas Kuehne wrote:

 -----BEGIN PGP SIGNED MESSAGE-----
 Hash: SHA1
 
 Anders F Björklund schrieb am Tue, 12 Apr 2005 15:56:25 +0200:
 

Now it seems to be disliking my paths instead:

i386-redhat-linux-gcc -o internal/gc/gc.o -O2 -g -pipe -march=i386 
-funittest -nostdinc -I ../gcc-3.4.3/gcc/d/phobos -I 
../gcc-3.4.3/gcc/d/phobos/internal/gc -I ./i386-redhat-linux -c 
../gcc-3.4.3/gcc/d/phobos/internal/gc/gc.d
{standard input}: Assembler messages:
{standard input}:10: Error: unknown pseudo-op: `..'
make: *** [internal/gc/gc.o] Error 1


 
 
 guessing from the error message your directory setup would be
 
 somewhere/gcc-3.4.3
 somewhere/gcc-3.4.3/gcc/d/phobos
 somewhere/phobos [you are here and trying to build Phobos]
 
 Is that right?
 
 Thomas
 
 
 -----BEGIN PGP SIGNATURE-----
 
 iD8DBQFCW9j93w+/yD4P9tIRAtdVAKCaxfmIifnbwZTuKdQGSIf4SEXNwACfXCBl
 9YM/Vgdo3SSoVv4rT2Q8rRQ=
 =Ujhp
 -----END PGP SIGNATURE-----


Argh, I forgot to mention this: The build process has changed and there 
is no need to build Phobos separately.  The INSTALL file has been 
updated and explains the new process. Don't worry, it's a lot easier now.

David
Apr 12 2005
parent =?ISO-8859-1?Q?Anders_F_Bj=F6rklund?= <afb algonet.se> writes:
David Friedman wrote:


 Argh, I forgot to mention this: The build process has changed and there 
 is no need to build Phobos separately.  The INSTALL file has been 
 updated and explains the new process. Don't worry, it's a lot easier now.


Yeah, ended up actually *reading* the INSTALL when the build failed :-)

And with a little luck, it should be no longer needed to provide a
/opt/gdc/gcc and /opt/gdc/g++ for Mac OS X... That should cut down
the download size for GDC, from 30 MB to a tiny 3 MB or something!

The vanilla build completed on Mac OS X, so I will test that later on...

--anders
Apr 12 2005
prev sibling parent =?ISO-8859-1?Q?Anders_F_Bj=F6rklund?= <afb algonet.se> writes:
Thomas Kuehne wrote:



 guessing from the error message your directory setup would be
 
 somewhere/gcc-3.4.3
 somewhere/gcc-3.4.3/gcc/d/phobos
 somewhere/phobos [you are here and trying to build Phobos]
 
 Is that right?


Actually my set up is: (done by the RPM specfile, you see...)

BUILD/
   gdc-0.11pre1-gcc-3.4.3/
     gcc-3.4.3
     obj-i386-redhat-linux
     gdc-temp
     phobos-unittest
     phobos-contracts
     phobos-release

Where gcc-3.4.3 are the patched GCC/GDC sources, unpacked from 
gcc-core-%{gcc_version}.tar.bz2 and gcc-g++-%{gcc_version}.tar.bz2,
and then gdc-%{gdc_version}.tar.bz2 is added as "gcc-?.?.?/gcc/d"

And I'm building GCC/GDC in the "obj-%{_target_platform}" directory,
and temporarily installing it in the "gdc-temp" directory to build
Phobos in the "phobos-***" and the "phobos-release" directories...


But I will fall back to building a vanilla version, since it seems that
David has added a new toplev patch and that my Make hacks don't work:

make libphobos.a unittest \
CC="gcc" CFLAGS="$OPT_FLAGS -DPHOBOS_USE_PTHREADS=1 -DHAVE_CONFIG_H 
-Ietc/c/stlsoft -I." \
DMD="gdc" DFLAGS="$OPT_FLAGS -funittest -nostdinc"

./unittest || exit 1

It's possible that this can be made simpler now, with the new
"check-target-libphobos"

Maybe one can even pass the flags directly to configure, now
that the whining "boehm-gc" target is not needed anymore ?

Who knows, maybe it can even be built with "make -j3" now...


All that I *really* want to do is to pass $OPT_FLAGS along
to the various C/C++/D builds - it is currently constructed from
$RPM_OPT_FLAGS, minus all flags that GCC and libiberty hates:

OPT_FLAGS=`echo $RPM_OPT_FLAGS|sed -e 's/-fno-rtti//g' -e 
's/-fno-exceptions//g'`

OPT_FLAGS=`echo $OPT_FLAGS|perl -pe 's/ ?-mcpu=([^ ]+)//g'`

Where the standard i386 $RPM_OPT_FLAGS is:
#rpm --eval %{optflags}
-O2 -g -pipe -march=i386 -mcpu=i686

Then I use the setting of "-funittest" for phobos-unittest,
"" for phobos-contracts and "-frelease" for phobos-release ?

Rest of the flags crud was just added to make it limp along...

--anders
Apr 12 2005
prev sibling next sibling parent =?ISO-8859-1?Q?Anders_F_Bj=F6rklund?= <afb algonet.se> writes:
David Friedman wrote:


 Lots of new features, not much documentation.  The full release will be 
 made when I sync up with the latest DMD version.


I get this error when trying to build Phobos:
"configure: error: Sorry, you must provide the
target-specific configure fragments (--enable-phobos-config-dir)"

I tried adding --disable-phobos-config-dir,
but that didn't help much... And making it a
fake directory just made "frag-unix not found".

--anders
Apr 12 2005
prev sibling next sibling parent =?ISO-8859-1?Q?Anders_F_Bj=F6rklund?= <afb algonet.se> writes:
David Friedman wrote:


 * Included Anders Björklund's patches to enable the use of frameworks on 
 Darwin.


In case anyone wonders, the patch is just to make GCC/GDC pass any
-framework on to the system linker, to avoid the clunky alternative:

	"-framework Whatever"  =>  "-Wl,-framework,Whatever"

It just makes doing GDC things a little nicer on the Mac OS X platform.
Especially with the new fix to make it co-exist with Apple's GCC/G++ !

--anders
Apr 12 2005
prev sibling next sibling parent reply Thomas Kuehne <thomas-dloop kuehne.thisisspam.cn> writes:
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

David Friedman schrieb am Tue, 12 Apr 2005 00:03:40 -0400:

 Lots of new features, not much documentation.  The full release will be 
 made when I sync up with the latest DMD version.

 http://home.earthlink.net/~dvdfrdmn/d/


FYI

http://dstress.kuehne.cn/raw_results/linux-i686_gdc-0.11pre
http://dstress.kuehne.cn/raw_results/linux-i686_gdc-0.11pre.log

Thomas


-----BEGIN PGP SIGNATURE-----

iD8DBQFCW+Sy3w+/yD4P9tIRAqvJAJ41YaZMHnkCl57GyKAVayE0IkPbFgCdH3Px
fW+phdVz3xZCSEUPi8gPUJY=
=nHdx
-----END PGP SIGNATURE-----
Apr 12 2005
parent reply Thomas Kuehne <thomas-dloop kuehne.thisisspam.cn> writes:
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1


 David Friedman schrieb am Tue, 12 Apr 2005 00:03:40 -0400:

 Lots of new features, not much documentation.  The full release will be 
 made when I sync up with the latest DMD version.

 http://home.earthlink.net/~dvdfrdmn/d/


 FYI

 http://dstress.kuehne.cn/raw_results/linux-i686_gdc-0.11pre
 http://dstress.kuehne.cn/raw_results/linux-i686_gdc-0.11pre.log


Thanks to Anders:
http://dstress.kuehne.cn/raw_results/mac-OS-X-10.3.7_gdc-0.11pre
http://dstress.kuehne.cn/raw_results/mac-OS-X-10.3.7_gdc-0.11pre.log

Thomas


-----BEGIN PGP SIGNATURE-----

iD8DBQFCW/z73w+/yD4P9tIRAtFRAKDHnRn0V/T1xL5k5B4tudvvMwhupACgu19o
HF0tc7TJeEYjA05VJeu8xVM=
=bU1P
-----END PGP SIGNATURE-----
Apr 12 2005
parent reply =?ISO-8859-1?Q?Anders_F_Bj=F6rklund?= <afb algonet.se> writes:
Thomas Kuehne wrote:


 Thanks to Anders:
 http://dstress.kuehne.cn/raw_results/mac-OS-X-10.3.7_gdc-0.11pre
 http://dstress.kuehne.cn/raw_results/mac-OS-X-10.3.7_gdc-0.11pre.log


Actually it was 10.3.8, but there's not really any real difference.

"Panther"... BTW, 10.4 Tiger - with GCC 4.0 - was announced today!

--anders
Apr 12 2005
parent David Friedman <d3rdclsmail_a_ _t_earthlink_d_._t_net> writes:
Anders F Björklund wrote:

 Thomas Kuehne wrote:
 

 Thanks to Anders:
 http://dstress.kuehne.cn/raw_results/mac-OS-X-10.3.7_gdc-0.11pre
 http://dstress.kuehne.cn/raw_results/mac-OS-X-10.3.7_gdc-0.11pre.log


 
 
 Actually it was 10.3.8, but there's not really any real difference.
 
 "Panther"... BTW, 10.4 Tiger - with GCC 4.0 - was announced today!
 
 --anders


  Tyger! Tyger! burning bright
  I'm shelling out 130 bucks tonight :-p
Apr 12 2005
prev sibling next sibling parent reply =?ISO-8859-1?Q?Anders_F_Bj=F6rklund?= <afb algonet.se> writes:
David Friedman wrote:


 Lots of new features, not much documentation.  The full release will be 
 made when I sync up with the latest DMD version.


I don't think you need to do that. You can release this as gdc 0.11 ?
And then release DMD 0.120 compatibility as gdc 0.12 "pre" instead...


 Major Improvements:
 
 * Included Anders Björklund's patches to enable the use of frameworks on 
 Darwin.


(and from the GDC home page: )


 Likewise, gdc can link C++ object code built by the system g++.


This is most *excellent* news! Now you can build a link D programs
with regular C libraries and system frameworks, without the need
for a custom GCC build installed on the system. Works just great!

The download/install also shrinked considerable, compared to 0.10:

Binaries:
8.0K    /opt/gdc/bin/dmd
288K    /opt/gdc/bin/gdc
2.0M    /opt/gdc/include/d
2.3M    /opt/gdc/lib/libphobos.a
4.0K    /opt/gdc/lib/libphobos.spec
3.1M    /opt/gdc/lib/gcc-lib/powerpc-apple-darwin7.8.0/3.3.5/cc1d
4.0K    /opt/gdc/lib/gcc-lib/powerpc-apple-darwin7.8.0/3.3.5/crt2.o
4.0K    /opt/gdc/share/man/man1/dmd.1
4.0K    /opt/gdc/share/man/man1/gdc.1
====
2.0M    gdc.zip

Sources:
  11M    gcc-core-3.3.5.tar.bz2
2.1M    gcc-g++-3.3.5.tar.bz2
788K    gdc-0.11pre1.tar.bz2

One might even consider "promoting" GDC back up to /usr again...
Since now it wouldn't interfere with the system C/C++ compilers.

But compiling D is now as simple as a /opt/gdc/bin/gdc, anyway.
D usability on the Mac OS X platform climbed up a whole ladder!


Now all we need is those D name-mangling patches for GDC,
but they seem to be coming along just nicely on Dsource ?

http://www.dsource.org/forums/viewforum.php?f=58
(demangler is done, "just" integration left todo)

--anders
Apr 12 2005
parent reply David Friedman <d3rdclsmail_a_ _t_earthlink_d_._t_net> writes:
Anders F Björklund wrote:

 David Friedman wrote:
 

 Lots of new features, not much documentation.  The full release will 
 be made when I sync up with the latest DMD version.


 
 
 I don't think you need to do that. You can release this as gdc 0.11 ?
 And then release DMD 0.120 compatibility as gdc 0.12 "pre" instead...
 


I still need to fix some bugs and update the documentation (including 
your man pages ;-)

David


 Major Improvements:

 * Included Anders Björklund's patches to enable the use of frameworks 
 on Darwin.


 
 
 (and from the GDC home page: )
 


 Likewise, gdc can link C++ object code built by the system g++.


 
 
 This is most *excellent* news! Now you can build a link D programs
 with regular C libraries and system frameworks, without the need
 for a custom GCC build installed on the system. Works just great!
 
 The download/install also shrinked considerable, compared to 0.10:
 
 Binaries:
 8.0K    /opt/gdc/bin/dmd
 288K    /opt/gdc/bin/gdc
 2.0M    /opt/gdc/include/d
 2.3M    /opt/gdc/lib/libphobos.a
 4.0K    /opt/gdc/lib/libphobos.spec
 3.1M    /opt/gdc/lib/gcc-lib/powerpc-apple-darwin7.8.0/3.3.5/cc1d
 4.0K    /opt/gdc/lib/gcc-lib/powerpc-apple-darwin7.8.0/3.3.5/crt2.o
 4.0K    /opt/gdc/share/man/man1/dmd.1
 4.0K    /opt/gdc/share/man/man1/gdc.1
 ====
 2.0M    gdc.zip
 
 Sources:
  11M    gcc-core-3.3.5.tar.bz2
 2.1M    gcc-g++-3.3.5.tar.bz2
 788K    gdc-0.11pre1.tar.bz2
 
 One might even consider "promoting" GDC back up to /usr again...
 Since now it wouldn't interfere with the system C/C++ compilers.
 
 But compiling D is now as simple as a /opt/gdc/bin/gdc, anyway.
 D usability on the Mac OS X platform climbed up a whole ladder!
 
 
 Now all we need is those D name-mangling patches for GDC,
 but they seem to be coming along just nicely on Dsource ?
 
 http://www.dsource.org/forums/viewforum.php?f=58
 (demangler is done, "just" integration left todo)
 
 --anders
Apr 12 2005
parent =?ISO-8859-1?Q?Anders_F_Bj=F6rklund?= <afb algonet.se> writes:
David Friedman wrote:



 I don't think you need to do that. You can release this as gdc 0.11 ?
 And then release DMD 0.120 compatibility as gdc 0.12 "pre" instead...


 
 I still need to fix some bugs and update the documentation (including 
 your man pages ;-)


I meant "fix the bugs in this one, *then* release it as 0.11" :-)

And wait with the new DMD features and stuff for 0.12 instead ?
(Assuming that it would be easier to release 0.11 without them)

--anders



 Apr 12 2005



prev sibling next sibling parent reply "Charlie" <charles jwavro.com>  writes:


A mingwn build soon to follow :) ?

Charlie

"David Friedman" <d3rdclsmail_a_ _t_earthlink_d_._t_net> wrote in message
news:d3fh9n$1gur$1 digitaldaemon.com...

 Lots of new features, not much documentation.  The full release will be
 made when I sync up with the latest DMD version.

 http://home.earthlink.net/~dvdfrdmn/d/

 Major Improvements:

 * Updated to DMD 0.113
 * Phobos is now built as target library (i.e., no need for a separate
 build step.)
 * Boehm-gc is no longer used and the Java package is no longer required.
 * Inline assembler for x86 (some limitations compared to DMD)
 * Included Anders Björklund's patches to enable the use of frameworks on
 Darwin.

 David





 Apr 12 2005



 next sibling parent reply bug d.com  writes:


Tried the pre-release with my example large file. cc1d reports:

virtual memory exhausted: Cannot allocate memory

Even though I have ~350M phisical free memory as shown by top.  gdc fails after
about 250M allocation.

What's the flag I should pass to "configure" to build a debugging version of
gdc?  I can try to generate the stack trace of gdc itself when it fails.  Do you
think this will help?



 Apr 12 2005



  parent reply David Friedman <d3rdclsmail_a_ _t_earthlink_d_._t_net>  writes:


bug d.com wrote:

 Tried the pre-release with my example large file. cc1d reports:
 
 virtual memory exhausted: Cannot allocate memory
 
 Even though I have ~350M phisical free memory as shown by top.  gdc fails after
 about 250M allocation.
 
 What's the flag I should pass to "configure" to build a debugging version of
 gdc?  I can try to generate the stack trace of gdc itself when it fails.  Do
you
 think this will help?
 
 
 


A stack trace will definately be helpful.

Setting the CFLAGS environment variable to '-O0 -g' before running 
configure should work.  You could also uncomment 'D_CC_FLAGS += -O0 -g' 
in d/Make-lang.in.

David



 Apr 12 2005



  parent reply bug d.com  writes:


This problem also doesn't work under gdc:

http://www.digitalmars.com/drn-bin/wwwnews?digitalmars.D/21401

Some info regarding my bug with large files: I have lot's of things like (T/A/B)
as described in the above.

Separate compilation always gives many errors like:

b.d:9: size of type T!(B) is not known
b.d:9: size of type T!(B) is not known

I'm not sure if this is a bug or feature? (both DMD and GDC behave the same.)

So I wrote a script to generate all of them into one big file, then DMD works.
But GDC fails with: "virtual memory exhausted: Cannot allocate memory".

By setting CFLAGS='-O0 -g', I was able to debug gdc, but I don't know where to
set breakpoint.

Detaching after fork from child process 20376.
virtual memory exhausted: Cannot allocate memory

Program exited with code 01.
(gdb)

Grep the error message doesn't give me much help, there are many occurance of
"virtual memory exhausted", but not occurance of "Cannot allocate memory".

Can you give me some hint here, where to set break point to catch the stack
trace?



 Apr 13 2005



  parent reply David Friedman <d3rdclsmail_a_ _t_earthlink_d_._t_net>  writes:


bug d.com wrote:

 This problem also doesn't work under gdc:
 
 http://www.digitalmars.com/drn-bin/wwwnews?digitalmars.D/21401
 
 Some info regarding my bug with large files: I have lot's of things like
(T/A/B)
 as described in the above.
 
 Separate compilation always gives many errors like:
 
 b.d:9: size of type T!(B) is not known
 b.d:9: size of type T!(B) is not known
 
 I'm not sure if this is a bug or feature? (both DMD and GDC behave the same.)
 
 So I wrote a script to generate all of them into one big file, then DMD works.
 But GDC fails with: "virtual memory exhausted: Cannot allocate memory".
 
 By setting CFLAGS='-O0 -g', I was able to debug gdc, but I don't know where to
 set breakpoint.
 
 Detaching after fork from child process 20376.
 virtual memory exhausted: Cannot allocate memory
 
 Program exited with code 01.
 (gdb)
 
 Grep the error message doesn't give me much help, there are many occurance of
 "virtual memory exhausted", but not occurance of "Cannot allocate memory".
 
 Can you give me some hint here, where to set break point to catch the stack
 trace?
 
 


The program to run in gdb is 'cc1d'.  You can run gdc with the '-v' 
option to see what the exact command line for cc1d is.  Then set 
breakpoints on 'exit' and 'perror'.

David



 Apr 13 2005



  parent reply bug d.com  writes:


In article <d3kfqs$2ptg$1 digitaldaemon.com>, David Friedman says...

The program to run in gdb is 'cc1d'.  You can run gdc with the '-v' 
option to see what the exact command line for cc1d is.  Then set 
breakpoints on 'exit' and 'perror'.


Ah, I tried to set BP on 'exit' and 'perror' before with gdc, now I know why it
doesn't work :-)

Here is the stack trace I got by watching cc1d.  If you need any more info, let
me know.  Thanks.

-------------------------------------------------------------------------

(gdb) b perror
Breakpoint 2 at 0x4015f256
(gdb) b exit
Breakpoint 3 at 0x4013a386
(gdb) cont
Continuing.
GNU D version 3.3.5 (i686-pc-linux-gnu)
compiled by GNU C version 3.3.3 20040412 (Gentoo Linux 3.3.3-r6, ssp-3.3.2-2,
pie-8.7.6).
GGC heuristics: --param ggc-min-expand=47 --param ggc-min-heapsize=32768

cc1d: out of memory allocating 4072 bytes after a total of 199913472 bytes

Breakpoint 3, 0x4013a386 in exit () from /lib/libc.so.6
(gdb) where









at d/d-glue.cc:1967

at d/dmd/attrib.c:129

at d/dmd/toobj.c:211

at d/dmd/template.c:1829






(gdb)



 Apr 13 2005



  parent reply David Friedman <d3rdclsmail_a_ _t_earthlink_d_._t_net>  writes:


bug d.com wrote:

 In article <d3kfqs$2ptg$1 digitaldaemon.com>, David Friedman says...
 

The program to run in gdb is 'cc1d'.  You can run gdc with the '-v' 
option to see what the exact command line for cc1d is.  Then set 
breakpoints on 'exit' and 'perror'.


 
 
 Ah, I tried to set BP on 'exit' and 'perror' before with gdc, now I know why it
 doesn't work :-)
 
 Here is the stack trace I got by watching cc1d.  If you need any more info, let
 me know.  Thanks.
 
 -------------------------------------------------------------------------
 
 (gdb) b perror
 Breakpoint 2 at 0x4015f256
 (gdb) b exit
 Breakpoint 3 at 0x4013a386
 (gdb) cont
 Continuing.
 GNU D version 3.3.5 (i686-pc-linux-gnu)
 compiled by GNU C version 3.3.3 20040412 (Gentoo Linux 3.3.3-r6, ssp-3.3.2-2,
 pie-8.7.6).
 GGC heuristics: --param ggc-min-expand=47 --param ggc-min-heapsize=32768
 
 cc1d: out of memory allocating 4072 bytes after a total of 199913472 bytes
 
 Breakpoint 3, 0x4013a386 in exit () from /lib/libc.so.6
 (gdb) where









 at d/d-glue.cc:1967

 at d/dmd/attrib.c:129

 at d/dmd/toobj.c:211

 at d/dmd/template.c:1829






 (gdb)
 
 


Here's something else to try:  Apply this patch to gcc/ggc-page.c and 
rebuild.  There's no need to run it in gdb; just try compiling the 
source file.  If it works, it means I have a garbage collection bug.

David



 Apr 24 2005



  parent  bug d.com  writes:


Tried, the same problem:

virtual memory exhausted: Cannot allocate memory

But the line number is different, I'm using gcc-3.3.5.



Here's something else to try:  Apply this patch to gcc/ggc-page.c and 
rebuild.  There's no need to run it in gdb; just try compiling the 
source file.  If it works, it means I have a garbage collection bug.

David


--------------010200080606090909070300
Content-Type: text/plain; x-mac-type="0"; x-mac-creator="0";
 name="gcc-page.patch"
Content-Transfer-Encoding: 7bit
Content-Disposition: inline;
 filename="gcc-page.patch"

*** ggc-page.c.orig	Tue Apr 19 19:14:34 2005
--- ggc-page.c	Sun Apr 24 15:24:37 2005
***************
*** 1774,1779 ****
--- 1774,1780 ----
  void
  ggc_collect (void)
  {
+     return;
    /* Avoid frequent unnecessary work by skipping collection if the
       total allocations haven't expanded much since the last
       collection.  */

--------------010200080606090909070300--





 Apr 24 2005



prev sibling  parent  David Friedman <d3rdclsmail_a_ _t_earthlink_d_._t_net>  writes:


Because this was a pre-release and there are still some outstanding 
issues, I didn't want to announce it just yet... But yes, you can at 
least build a cross compiler with a mingw32 target now.  I haven't tried 
a native build under MSYS/Cygwin.

BTW, this is the only cross target currently supported 'out of the box'.

David

Charlie wrote:

 A mingwn build soon to follow :) ?
 
 Charlie
 
 "David Friedman" <d3rdclsmail_a_ _t_earthlink_d_._t_net> wrote in message
 news:d3fh9n$1gur$1 digitaldaemon.com...
 

Lots of new features, not much documentation.  The full release will be
made when I sync up with the latest DMD version.

http://home.earthlink.net/~dvdfrdmn/d/

Major Improvements:

* Updated to DMD 0.113
* Phobos is now built as target library (i.e., no need for a separate
build step.)
* Boehm-gc is no longer used and the Java package is no longer required.
* Inline assembler for x86 (some limitations compared to DMD)
* Included Anders Björklund's patches to enable the use of frameworks on
Darwin.

David


 
 
 





 Apr 12 2005



prev sibling  parent  brad domain.invalid  writes:


David Friedman wrote:

 Lots of new features, not much documentation.  The full release will be 
 made when I sync up with the latest DMD version.
 
 http://home.earthlink.net/~dvdfrdmn/d/
 
 Major Improvements:
 
 * Updated to DMD 0.113
 * Phobos is now built as target library (i.e., no need for a separate 
 build step.)
 * Boehm-gc is no longer used and the Java package is no longer required.
 * Inline assembler for x86 (some limitations compared to DMD)
 * Included Anders Björklund's patches to enable the use of frameworks on 
 Darwin.
 
 David


I just wanted to say thank you for all the work that you have put in to GDC.

Thanks
Brad



 Apr 13 2005