String concatenation, mingw, and the + operator

Lars Andersen
Fri Mar 28 08:40:00 GMT 2003

On Thu, 2003-03-27 at 19:59, Tom Tromey wrote:
> >>>>> "Lars" == Lars Andersen <> writes:
> Lars>     System.out.println("Test");
> Lars>     System.out.println(""+0);
> Lars>     System.out.println(""+1);
> Lars> , but this is what i get :
> Lars> Test
> Lars> 1
> I get this too :-(.

Just tested on GCC 3.2.1 and GCC 3.3 (latest snapshot) on x86 Linux.

The bug is not present in 3.2.1, but it is present in 3.3 snapshot.

I suspected that +0 get optimized away, like in an integer expression
(a+5+0), where you can safely remove the +0 part without altering the
result. Of course this doesn't work for string concatenations.

But compiling with -O0 -O1 -O2 all yields the same result.
Maybe its something different.

> We've had bugs in this area before.  I guess you just found another
> one.  Luckily this doesn't seem to happen much in real code.
> Could you submit a PR for this?
Will do.

> BTW, if anybody has been itching to fix a compiler bug, this one might
> be a good place to start.

Well, I will give it a shot then.

/Lars Andersen

More information about the Java mailing list