Re: gcj trunk on OS X

[Andreas Tobler <toa at pop dot agri dot ch>]

Hi Mohan,

Mohan Embar wrote:

> > > > What Xcode version do you have installed ?
> > > Xcode: 2.2.1
> > > Xcode IDE: 656.0
> > > Xcode Core: 660.0
> > > ToolSupport: 651.0
> > Aha! This the issue. If you download and install an Xcode 2.4.1 for 
> > example, you'd be able to use the below. These headers get an update 
> > while you install the new Xcode.
>
> Thanks for the tip. I don't mind upgrading.

Ok. Good.

> Another question, while I have you here.
>
> If I try to compile javolution.jar, which can be obtained
> from here:
>
> http://www.javolution.org/javolution-4.2.5-bin.zip
>
> (BSD license.)
>
> With my built gcj, I get this when I compile with -O2:
>
> mohanembar$ gcj -c -O2 /Data/temp/javolution/javolution-4.2/javolution.jar 
> javolution/lang/MathLib.java: In class 'javolution.lang.MathLib':
> javolution/lang/MathLib.java: In method 'javolution.lang.MathLib.toLongPow2(double,int)':
> javolution/lang/MathLib.java:0: internal compiler error: in build_int_cst_wide, at tree.c:885
> Please submit a full bug report,
> with preprocessed source if appropriate.
> See <URL:http://gcc.gnu.org/bugs.html> for instructions.
>
> This doesn't happen when I leave out the -O2.
>
> Does this seem familiar to you? I'll submit a Bugzilla
> report if you want me to.

No it doesn't. File a report please. (me is only a porter, know nothing 
about java, anyway I think it has to do with compiler internals and not 
directly with java.)

As a side note. At the moment the libjava build breaks under 
i686-apple-darwin. The last good revision is 122040.

So do not upgrade if you want to test real java code :)

I'll try to tackle the failure:

/usr/bin/ld: java/.libs/lang.o has external relocation entries in 
non-writable section (__TEXT,__text) for symbols:
java::lang::Class* java::lang::Class::forName(java::lang::String*, 
java::lang::Class*)

It is showing up since 122041, but I guess it has nothing to do with the 
patch itself. It only 'uncovers' an i686-apple-darwin stub problem....

Andreas