This looks like a runtime initialization problem. Class.forName(), and subsequently, VMClassLoader 
is being called from inside _Jv_CreateJavaVM, but the File.separator field has not yet been set.

1. Does a "hello world" program work ok? 

2. Can you try this with the current CVS code?

3. Can you narrow this problem down to a small test case?

Subject: Re:  SIGSEGV, Segmentation fault starting java
	program

On ke, 2003-10-22 at 01:38, bryce at mckinlay dot net dot nz wrote:
> PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12708
> ------- Additional Comments From bryce at mckinlay dot net dot nz  2003-10-21 22:38 -------
> This looks like a runtime initialization problem. Class.forName(), and subsequently, VMClassLoader 
> is being called from inside _Jv_CreateJavaVM, but the File.separator field has not yet been set.
> 
> 1. Does a "hello world" program work ok? 

No, same error. Good Question! I was stupid to not tested myself.

> 2. Can you try this with the current CVS code?

No, I tried, but I seems to have difficulties to compile it:
-I../../gcc/gcc/ada ../../gcc/gcc/ada/gnat1drv.adb -o ada/gnat1drv.o
gnatbind -C -I- -I. -Iada -I../../gcc/gcc/ada -o ada/b_gnat1.c -n
ada/gnat1drv.ali
fatal error: file gnat1drv.ali is incorrectly formatted
make sure you are using consistent versions of gcc/gnatbind
9.   R nnvvnnnnnnnnnvvvnnvnvvnnnnnnnnnvnvnnnnnnnvvvnnnnvvnn
                                                     |
make[2]: *** [ada/b_gnat1.c] Error 4
make[2]: Leaving directory `/tmp/pj/cvs/objdir/gcc'
make[1]: *** [stage1_build] Error 2
make[1]: Leaving directory `/tmp/pj/cvs/objdir/gcc'
make: *** [bootstrap] Error 2

I can try again with other configuration options if you like. 

> 3. Can you narrow this problem down to a small test case?

Yes!
 cat > HelloWorld.java
class HelloWorld {
    public static void main(String[] args) {
        System.out.println("Hello World!");
}
}
[pj@rubiini pj]$ /v/linux22_i386/lang/gcc/3.3.2/bin/gcj
--main=HelloWorld -g -o HelloWorld HelloWorld.java
[pj@rubiini pj]$ ./HelloWorld 
Väärä muistilohko (muistin tila talletettu)
oops my LANG was fi_FI
setenv LANG en_EN
./HelloWorld
Segmentation fault (core dumped)

I'm compiling gcc again with -march=i686 instead of pentium4 

Subject: Re:  SIGSEGV, Segmentation fault starting java program

>> 1. Does a "hello world" program work ok?
>
> No, same error. Good Question! I was stupid to not tested myself.

OK, so it must be something caused by your configuration or 
environment. What linux distribution is this on? And are you using the 
distribution's gcj or did you compile it yourself?

>> 2. Can you try this with the current CVS code?
>
> No, I tried, but I seems to have difficulties to compile it:
> -I../../gcc/gcc/ada ../../gcc/gcc/ada/gnat1drv.adb -o ada/gnat1drv.o
> gnatbind -C -I- -I. -Iada -I../../gcc/gcc/ada -o ada/b_gnat1.c -n
> ada/gnat1drv.ali
> fatal error: file gnat1drv.ali is incorrectly formatted
> make sure you are using consistent versions of gcc/gnatbind

Try configuring with --enable-languages=java - this will only build the 
C, C++, and java compilers so problems with ada will be avoided (and 
the build will go a bit faster)

FYI my configure command is:

./configure --prefix=/home/bryce/gcc --enable-languages=java 
--enable-java-awt=gtk --disable-static

Subject: Re:  SIGSEGV, Segmentation fault starting java
	program

On ke, 2003-10-22 at 11:29, bryce at mckinlay dot net dot nz wrote:
> PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12708

> OK, so it must be something caused by your configuration or 
> environment. What linux distribution is this on? 

It's RedHat 9 with all updates:
uname -a
Linux rubiini.csc.fi 2.4.20-20.9smp #1 SMP Mon Aug 18 11:32:15 EDT 2003
i686 i686 i386 GNU/Linux

> And are you using the 
> distribution's gcj or did you compile it yourself?

I compiled myself. I'll try cvs again.

Subject: Re:  SIGSEGV, Segmentation fault starting java
	program

On ke, 2003-10-22 at 11:29, bryce at mckinlay dot net dot nz wrote:
> Try configuring with --enable-languages=java - this will only build the 
> C, C++, and java compilers so problems with ada will be avoided (and 
> the build will go a bit faster)

It helps but:

../../../gcc/libiberty/fibheap.c: In function `fibheap_delete_node':
 
../../../gcc/libiberty/fibheap.c:285: error: `LONG_MIN' undeclared
(first use in this function)
../../../gcc/libiberty/fibheap.c:285: error: (Each undeclared identifier
is reported only once
../../../gcc/libiberty/fibheap.c:285: error: for each function it
appears in.)
make[2]: *** [fibheap.o] Error 1
make[2]: Leaving directory
`/tmp/pj/cvs/objdir/i686-pc-linux-gnu/libiberty'
make[1]: *** [all-target-libiberty] Error 2
make[1]: Leaving directory `/tmp/pj/cvs/objdir'
make: *** [bootstrap] Error 2

Subject: Re:  SIGSEGV, Segmentation fault starting java program

On Oct 22, 2003, at 11:49 PM, pj at csc dot fi wrote:

> ../../../gcc/libiberty/fibheap.c: In function `fibheap_delete_node':
>
> ../../../gcc/libiberty/fibheap.c:285: error: `LONG_MIN' undeclared
> (first use in this function)
> ../../../gcc/libiberty/fibheap.c:285: error: (Each undeclared 
> identifier
> is reported only once
> ../../../gcc/libiberty/fibheap.c:285: error: for each function it
> appears in.)
> make[2]: *** [fibheap.o] Error 1

Well, something is very wrong here - looks like a problem with your 
system/glibc headers?

Subject: Re:  SIGSEGV, Segmentation fault starting java
	program

On ke, 2003-10-22 at 13:59, bryce at mckinlay dot net dot nz wrote:

> Well, something is very wrong here - looks like a problem with your 
> system/glibc headers?

Not very wrong because I can compile 3.3.2 (but it don't work with java)
The compilation problem is only with cvs version.
There can be something wrong, I'll try other computer, which is upgraded
to 9 from RedHat 8. 

Subject: Re:  SIGSEGV, Segmentation fault starting java
	program

On ke, 2003-10-22 at 13:59, bryce at mckinlay dot net dot nz wrote:

> Well, something is very wrong here - looks like a problem with your 
> system/glibc headers?

No, if I compile gcc 3.3.2 by command
make CFLAGS='-O2' LIBCFLAGS='-g -O2' LIBCXXFLAGS='-g -O2
-fno-implicit-templates' bootstrap 
HelloWorld works well but if I use
make CFLAGS='-O3 -march=i686 -mcpu=pentium4' LIBCFLAGS='-g -O3
-march=i686 -mcpu=pentium4' LIBCXXFLAGS='-g -O3 -march=i686 
-mcpu=pentium4 -fno-implicit-templates' bootstrap
it failes to start.

cvs version failed to compile also other Redhat 9 machine.

What version of binutils do you have?
From <http://gcc.gnu.org/install/specific.html#ix86-*-linux*>:
As of GCC 3.3, binutils 2.13.1 or later is required for this platform. See bug 10877 for more 
information.

Subject: Re:  SIGSEGV, Segmentation fault starting java program

RHL 9 ships with binutils 2.13.90 so I doubt this would be a problem.

Pekka, regarding your build problem, I build CVS GCC at least every few 
days on an RHL 9 system (and Im sure many others do), so the build 
error must have something to do with your environment or the commands 
you are using to build GCC.

FYI I build gcc using

cd cvs/gcc
mkdir build
cd build
../configure --prefix=/home/bryce/gcc --enable-languages=java 
--enable-java-awt=gtk --disable-static
make

The compiler you are using to build/bootstrap can also cause problems, 
make sure you are using redhat's compiler or a known-good GCC 3.3.x 
build in your path.

*** This bug has been marked as a duplicate of 10877 ***

reopen this bug because...

it's a duplicate of a different bug.

*** This bug has been marked as a duplicate of 13334 ***

This Driver version is error free,DataDirect free trial downloads for JDBC drivers, ODBC drivers and ADO.NET providers including connection to Oracle, DB2, Sybase, and MS SQL Server. Download it from: http://www.datadirect.com/products/jdbc/index.html