review request (S): 7199082 write warning messages to stderr

[John Coomes]

David Holmes (david.holmes at oracle.com) wrote:
> Hi John,
> 
> On 19/09/2012 4:47 AM, John Coomes wrote:
> > Small change to warning() to write to error_stream() instead of tty.
> >
> > http://cr.openjdk.java.net/~jcoomes/7199082-warn-stderr/
> 
> I still think treating warnings differently to other VM output is not 
> necessarily the right way to go here.

I think there is a difference.  You don't ask for warnings, but you do
ask for most other JVM output.  And warnings can be issued for things
that are not under the application's control - failure to allocate
shared memory, failure to allocate large pages, and so on.

> Further I'm concerned that the real problem here is that we shouldn't be 
> issuing warnings for the now obsolete PermGen flags. Whether the warning 
> goes to stderr or stdout it is new noise that might impact people's 
> development/build environments.

The warnings serve the purpose of noting a behavior change.  That
said, I'm not fully convinced that we need warnings for these
particular flags, but the JVM can issue warnings for any number of
reasons, and those should not interfere with the application output.
At the very least, I think the change is a step in the right
direction.

> So while this fix will address the immediate javac problem, and can be 
> considered Reviewed, I suspect we will have to revisit this.

Thanks.  Fortunately, there already is a plan to revisit it:

	http://openjdk.java.net/jeps/158

-John

> > Bug link:
> > http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=7199082
> >
> > -John