mingw32 gcj compile failure and possible patch

mingw32 gcj compile failure and possible patch

[aking]

This is my first attempted patch, so please be gentle :)
Without the attached (timy) patch, I am unable to compile and execute a  
simple 'hello world' program with the mingw32 port.  Unfortunately, I am 
much more of a java coder than a C programmer, so I'm not sure if it's a 
correct patch or not.  

Here's the gcj output without the patch:

aking@nanostorm:~/gcj-test$ i686-pc-mingw32-gcj --main=HelloWorld -o hello.exe
HelloWorld.java
/home/aking/mingw-gcc/lib/gcc-lib/i686-pc-mingw32/3.1/../../../../i686-pc-mingw32/lib/libgcj.a(win32.o):
In function `Z26_Jv_platform_close_on_execi':
/home/aking/compile/mingw-gcj/bin/i686-pc-mingw32/libjava/include/platform.h:25:
multiple definition of `_Jv_platform_close_on_exec(int)'
/home/aking/mingw-gcc/lib/gcc-lib/i686-pc-mingw32/3.1/../../../../i686-pc-mingw32/lib/libgcj.a(prims.o)(.text+0x0):/home/aking/compile/mingw-gcj/bin/i686-pc-mingw32/libjava/../../../gcc/libjava/prims.cc:
first defined here
/home/aking/mingw-gcc/lib/gcc-lib/i686-pc-mingw32/3.1/../../../../i686-pc-mingw32/lib/libgcj.a(natSystem.o):
In function
`Z34_Jv_GetArrayElementFromElementTypePN4java4lang6ObjectEPNS0_5ClassE':
/home/aking/compile/mingw-gcj/bin/i686-pc-mingw32/libjava/../../../gcc/libjava/gcj/cni.h(.text+0x0):
multiple definition of `_Jv_platform_close_on_exec(int)'
/home/aking/mingw-gcc/lib/gcc-lib/i686-pc-mingw32/3.1/../../../../i686-pc-mingw32/lib/libgcj.a(prims.o)(.text+0x0):/home/aking/compile/mingw-gcj/bin/i686-pc-mingw32/libjava/../../../gcc/libjava/prims.cc:
first defined here
/home/aking/mingw-gcc/lib/gcc-lib/i686-pc-mingw32/3.1/../../../../i686-pc-mingw32/lib/libgcj.a(natPlainSocketImpl.o):
In function `Z26_Jv_platform_close_on_execi':
/home/aking/compile/mingw-gcj/bin/i686-pc-mingw32/libjava/include/platform.h:25:
multiple definition of `_Jv_platform_close_on_exec(int)'
/home/aking/mingw-gcc/lib/gcc-lib/i686-pc-mingw32/3.1/../../../../i686-pc-mingw32/lib/libgcj.a(prims.o)(.text+0x0):/home/aking/compile/mingw-gcj/bin/i686-pc-mingw32/libjava/../../../gcc/libjava/prims.cc:
first defined here
aking@nanostorm:~/gcj-test$

With the patch, the java program compiles and runs fine under Windows ME.

The patch:

*** gcc-3.1-branch/libjava/include/win32.h	Sun Mar 10 13:00:04 2002
--- mingw-gcj/gcc/libjava/include/win32.h	Mon Apr  1 22:21:36 2002
*************** details.  */
*** 21,27 ****
  extern void _Jv_platform_initialize (void);
  extern jlong _Jv_platform_gettimeofday ();
  
! void _Jv_platform_close_on_exec (jint)
  {
    // Ignore.
  }
--- 21,27 ----
  extern void _Jv_platform_initialize (void);
  extern jlong _Jv_platform_gettimeofday ();
  
! inline void _Jv_platform_close_on_exec (jint)
  {
    // Ignore.
  }

Re: mingw32 gcj compile failure and possible patch

[Adam Megacz]

Thank you for the patch! I took the opportunity to clean up the
formatting on that function, and will apply it as shown below
(branch+trunk) as soon as it finishes building.

  - a

Index: win32.h
===================================================================
RCS file: /cvs/gcc/gcc/libjava/include/win32.h,v
retrieving revision 1.3.8.2
diff -u -r1.3.8.2 win32.h
--- win32.h2002/03/10 18:00:041.3.8.2
+++ win32.h2002/04/02 18:48:24
@@ -21,7 +21,8 @@
 extern void _Jv_platform_initialize (void);
 extern jlong _Jv_platform_gettimeofday ();

-void _Jv_platform_close_on_exec (jint)
+inline void
+_Jv_platform_close_on_exec (jint)
 {
   // Ignore.
 }

-- 
The web is dead; long live the Internet.
http://www.xwt.org/

Re: mingw32 gcj compile failure and possible patch

[Frank Jacobs]

Thanks for sending out the patch! I finally got a GCJ
Linux->Win32 cross compiler working last night; it
almost brought tears to my eyes. ;-)

I have a question about executable size. When I tried
the standard Hello, World application, my resulting
executable was 10 MB. Any ideas as to why this may be?
I've read through the Cross Compiler FAQ, but couldn't
find any hints on this.

FYI, I configured and built the compiler using a
script posted by Adam in January. I modified it to
remove the patches in it, but it is basically the
same. See
http://gcc.gnu.org/ml/java/2002-01/msg00454.html for
Adam's original script.

Thanks a bunch -- I'm very much a beginner when it
comes to cross compilers, so my apologies if there is
an easy answer to this one.

--- Adam Megacz <gcj@lists.megacz.com> wrote:
> 
> Thank you for the patch! I took the opportunity to
> clean up the
> formatting on that function, and will apply it as
> shown below
> (branch+trunk) as soon as it finishes building.
> 
>   - a
> 
> Index: win32.h
>
===================================================================
> RCS file: /cvs/gcc/gcc/libjava/include/win32.h,v
> retrieving revision 1.3.8.2
> diff -u -r1.3.8.2 win32.h
> --- win32.h2002/03/10 18:00:041.3.8.2
> +++ win32.h2002/04/02 18:48:24
> @@ -21,7 +21,8 @@
>  extern void _Jv_platform_initialize (void);
>  extern jlong _Jv_platform_gettimeofday ();
> 
> -void _Jv_platform_close_on_exec (jint)
> +inline void
> +_Jv_platform_close_on_exec (jint)
>  {
>    // Ignore.
>  }
> 
> -- 
> The web is dead; long live the Internet.
> http://www.xwt.org/


__________________________________________________
Do You Yahoo!?
Yahoo! Tax Center - online filing with TurboTax
http://taxes.yahoo.com/

Re: mingw32 gcj compile failure and possible patch

[aking]

On Wed, Apr 03, 2002 at 10:23:11AM -0800, Frank Jacobs wrote:
> I have a question about executable size. When I tried
> the standard Hello, World application, my resulting
> executable was 10 MB. 

Just strip it:

aking@nanostorm:~/gcj-test$ i686-pc-mingw32-gcj --main=HelloWorld -o hello.exe HelloWorld.java
aking@nanostorm:~/gcj-test$ ls -al hello.exe
-rwxr-xr-x    1 aking    users    10824061 Apr  3 14:17 hello.exe*
aking@nanostorm:~/gcj-test$ strip hello.exe
aking@nanostorm:~/gcj-test$ ls -al hello.exe
-rwxr-xr-x    1 aking    users     1208320 Apr  3 14:17 hello.exe*

There now seems to be an exception when compiling java code that uses
exceptions:

aking@nanostorm:~/gcj-test$ i686-pc-mingw32-gcj --main=GCJTest -o gcjtest GCJTest.java
/home/aking/mingw-gcc/lib/gcc-lib/i686-pc-mingw32/3.1/../../../../i686-pc-mingw32/lib/libgcj.a(exception.o): In function `abort':
/home/aking/compile/mingw-gcj/bin/i686-pc-mingw32/libjava/../../../gcc/libjava/exception.cc:27: multiple definition of `abort'
/home/aking/mingw-gcc/lib/gcc-lib/i686-pc-mingw32/3.1/../../../../i686-pc-mingw32/lib/libmsvcrt.a(ds00510.o)(.text+0x0): first defined here

I was hoping Adam M. would have a chance to look at this error, but
if not, I'll try and track it down this evening.  The java code
is at the end.

	Adam K


public class GCJTest
{
	public static void main( String[] args )
	{
		try
		{
			Thread.sleep( 2000 );
		}
		catch( Exception e )
		{
		}
	}
}

Re: mingw32 gcj compile failure and possible patch

[Adam Megacz]

Frank Jacobs <forkjake@yahoo.com> writes:
> I have a question about executable size. When I tried the standard
> Hello, World application, my resulting executable was 10 MB.

If you strip it, it should go down to 3MB.

I'll be releasing gump (GCJ Unreachable Method Pruner) in a few weeks,
which should bring it down below a meg (209kb gzipped is my record so
far, but I know I can do even better than that).

  - a

-- 
The web is dead; long live the Internet.
http://www.xwt.org/

Re: mingw32 gcj compile failure and possible patch

[Frank Jacobs]

gump -- that'll be really cool!!!

--- Adam Megacz <gcj@lists.megacz.com> wrote:
> 
> Frank Jacobs <forkjake@yahoo.com> writes:
> > I have a question about executable size. When I
> tried the standard
> > Hello, World application, my resulting executable
> was 10 MB.
> 
> If you strip it, it should go down to 3MB.
> 
> I'll be releasing gump (GCJ Unreachable Method
> Pruner) in a few weeks,
> which should bring it down below a meg (209kb
> gzipped is my record so
> far, but I know I can do even better than that).
> 
>   - a
> 
> -- 
> The web is dead; long live the Internet.
> http://www.xwt.org/


__________________________________________________
Do You Yahoo!?
Yahoo! Tax Center - online filing with TurboTax
http://taxes.yahoo.com/

Re: mingw32 gcj compile failure and possible patch

[Adam Megacz]

aking@dreammechanics.com writes:
> I was hoping Adam M. would have a chance to look at this error, but

It should be fixed now -- could you try again?

2002-04-04  Adam Megacz <adam@xwt.org>

        * include/win32.h (_Jv_platform_close_on_exec): added inline
        modifier.

  - a

-- 
The web is dead; long live the Internet.
http://www.xwt.org/

Re: mingw32 gcj compile failure and possible patch

[Adam Megacz]

Adam Megacz <gcj@lists.megacz.com> writes:
> aking@dreammechanics.com writes:
> > I was hoping Adam M. would have a chance to look at this error, but

> It should be fixed now -- could you try again?

Er, that fix was for a differnt problem. Here's the fix for your
problem. Could you cvs update and try again?

  - a

2002-04-05  Adam Megacz <adam@xwt.org>

        * exception.cc (abort): added static modifier

Index: exception.cc
===================================================================
RCS file: /cvs/gcc/gcc/libjava/exception.cc,v
retrieving revision 1.20
diff -u -r1.20 exception.cc
--- exception.cc        2002/02/13 00:29:42     1.20
+++ exception.cc        2002/04/06 04:17:54
@@ -23,7 +23,8 @@
 // stdlib.h's abort().
 namespace std
 {
-  __attribute__ ((__noreturn__)) void abort ()
+  static __attribute__ ((__noreturn__)) void
+  abort ()
   {
     ::abort ();
   }

Re: mingw32 gcj compile failure and possible patch

[aking]

On Fri, Apr 05, 2002 at 08:16:04PM -0800, Adam Megacz wrote:
> 
> Adam Megacz <gcj@lists.megacz.com> writes:
> > aking@dreammechanics.com writes:
> > > I was hoping Adam M. would have a chance to look at this error, but
> 
> > It should be fixed now -- could you try again?
> 
> Er, that fix was for a differnt problem. Here's the fix for your
> problem. Could you cvs update and try again?
> 

Strange.. I just did an update, but I'm not seeing this patch
in exception.cc.  Is the cvs server I'm using some kind of mirror?
It's subversions.gnu.org:/cvsroot/gcc

Anyways, I applied your patch manually and it does indeed
fix the link error.  Thanks!

	Adam

> 2002-04-05  Adam Megacz <adam@xwt.org>
> 
>         * exception.cc (abort): added static modifier
> 
> Index: exception.cc
> ===================================================================
> RCS file: /cvs/gcc/gcc/libjava/exception.cc,v
> retrieving revision 1.20
> diff -u -r1.20 exception.cc
> --- exception.cc        2002/02/13 00:29:42     1.20
> +++ exception.cc        2002/04/06 04:17:54
> @@ -23,7 +23,8 @@
>  // stdlib.h's abort().
>  namespace std
>  {
> -  __attribute__ ((__noreturn__)) void abort ()
> +  static __attribute__ ((__noreturn__)) void
> +  abort ()
>    {
>      ::abort ();
>    }

Re: mingw32 gcj compile failure and possible patch

[Bryce McKinlay]

aking@dreammechanics.com wrote:

>Strange.. I just did an update, but I'm not seeing this patch
>in exception.cc.  Is the cvs server I'm using some kind of mirror?
>It's subversions.gnu.org:/cvsroot/gcc
>

Yes, the GCC CVS on subversions is a mirror which is updated once or 
twice a day, I think. gcc.gnu.org is the primary repository but it tends 
to be a lot slower, so please use subversions for anonymous CVS if 
possible.

regards

Bryce.