Hi, Roman,
Roman Kennke wrote:
Hi Artem,
the idea you're proposing is fine, actually so fine that we have already
partly implemented it in the latest JDK6 update releases :) The fix
itself is also fine from my perspective.
Unfortunately, our implementation is slightly different - I'm not sure
if you have access to 6u10/6u12 sources, though - so if we integrate
your fix as proposed to OpenJDK, we'll have some troubles with upcoming
merge with 6u10/6u12 codebase... The only thing I'd suggest for you now
is to rewrite your fix using the same approach as we have.
Hrmpf. Is this fix in OpenJDK6 already? Then I should be able to grab
the stuff and see what I can do. OTOH, I probably don't care so much
about this then.
the changes I told about are not present in OpenJDK6 yet.
BTW, to save your (and our) efforts next time, it may be worth notifying
everybody from this dev group about what you're going to start working
on - probably, someone else has almost implemented this in other manner.
This is by no means not a requirement, I'm just trying to avoid
conflicts like this in the future.
Yeah, would be nice. I'm not sure how to do that though.
Speaking of this, I'd like to get rid of those many initIDs() methods in
public AWT classes some day, because I don't need them in the peer impls
that I have, but they force me to be implemented w/ empty stubs. That's
quite ugly IMO and can be avoided in most cases by doing the JNI ID
initialization in the peer code that's actually using them. I know that
there's the class un-/reloading problem to be thought about, but in most
cases that is not an issue (e.g. a XYZButtonPeer.initIDs() can safely
init Button fields, because the Button class is not unloaded as long as
the XYZButtonPeer class is unloaded, etc).
Right, initIDs look weird and should be moved to peer code when
possible. However there are some cases when method IDs are stored for
java.awt.* classes - see Java_java_awt_Font_initIDs() or
Java_java_awt_FontMetrics_initIDs() for example. These cases should be
carefully inspected and also moved to peer code.
Thanks,
Artem
/Roman
