[antlr-interest] How to report bugs ?

[mblp at mega.ist.utl.pt]

Heh, the double checked locking doesn't work in Java  because of the
definition of the Java Memory Model, or lack of restrition with
regards of out-of-order execution. In fact, out-of-order is so common
that modern cpu's abuse it to get even the slightest 1% performance
gain.

As for FileSystem, it referes to a native call.

    /**
     * Return the FileSystem object representing this platform's local
     * filesystem.
     */
    public static native FileSystem getFileSystem();

So, it should work the same on all VMs.

Anyway, what was the issue...? :P

On 3/28/07, Dave Cramer <davec at postgresintl.com> wrote:
>
> On 28-Mar-07, at 4:12 AM, Martin Probst wrote:
>
> >> I this case, I don't know whether the compiler will in-line the value
> >> that is in effect while it runs or whether it will refrain from in-
> >> lining
> >> a value for that field.
> >
> > Eh, do you really think the people at Sun are not smart enough to
> > realize that the File.separator field is system dependent and
> > implement that in a way that is not going to break? I mean, they go
> > on and implement a whole VM, platform independent language spec
> > etc., and then they forget this? And no one notices in nearly 10
> > years, up to this day?
> >
> Is this the same company that created a VM which does not honour the
> order of operations thus creating the lazy initializer singleton
> problem ?
>
> http://www.cs.umd.edu/~pugh/java/memoryModel/DoubleCheckedLocking.html
>
> For what it's worth Jean emailed me back and told me the bug was
> fixed. I don't know what the issue was ?
> > Also, the compiler doesn't execute the code it's compiling, so it
> > simply can't inline that field because it has no way to determine
> > it's value. Code that is static, like static initializers or the
> > static block, is supposed to be executed at class load time, if I'm
> > not mistaken.
> >
> > Martin
>
>