Mario,
Did you, or can you, share a test case (and any platform specific info
needed) to repro this?
-phil.
Mario Torre wrote:
Il giorno ven, 12/06/2009 alle 10.13 -0700, Phil Race ha scritto:
If I don't set the loop in the
> constructor, and don't check for null in the getter, I get NPE in
> various places,
Isn't that just a bug in (I guess) your SurfaceData subclass ?
-phil.
Hi Phil!
No, because this is with the "plain" OpenJDK and not with my
SurfaceData. This is because all the users of this field expect a valid
non null value, but if we don't set it in the constructor, only the
calls that happens after validate() will see a non null value. Having
this null check in checkFontInfo also makes the NPE disappear in the
Java2D demo, thus obtaining the same effect as initialising in the
constructor, but without exposing "this". It has to be checked if there
are other possible location where this variable is used non initialised,
but I doubt. It's not really robust though if we depend on this
initialisation in the constructor the way it is in my opinion. Not that
is a serious bug, and in fact we could even say it's not a bug because
the code works, but it's not correct in my point of view and looks easy
to fix, so... Just want to be sure that the proposed solution doesn't
add new bugs in some corner cases or doesn't add overhead, so who knows
the code better has the last word on that, of course.
Cheers,
Mario
