; Date: Wed Jan 25 2006
As Ray Gans noted, the Mustang Beta release is coming up soon. Current plans are to have two beta periods and to ship the final release in the Autumn.
As I noted last week, we very much want to know about bugs and regressions that we find. As I said, we do a lot of testing all through the Java organization. But no matter how much we test there's a crucial test that we cannot do. Namely, we can't test your application in your environment.
Like I said: "We try and put ourselves in your shoes, to think about what you might do with the features, and write tests from that viewpoint. But, try as we might, we are not you, we do not have your environment, we do not have your applications, we do not have your mindset, and we can only go so far." and there was more along those lines which I said.
The important thing about finding and reporting bugs is to do so as early as possible. For a bug fix to get into the release it helps a lot to know early about the bug, rather than at the last minute. The earlier we know about the bug, the more time we have to develop the fix, and test the result.
Our goal is always to release a high quality software development platform. Part of doing that is extensive testing. We have some periods of running the full test cycle, and will spend 2-4 weeks to run that full test cycle.
Probably the worst sort of bug is the regression. This is where we make a change, and something breaks that used to work. I'm sorry to say, that happens, and that it happens for every software organization. I think part of the cause is what I said earlier, we are not you, we do not have your applications, we do not have your environment, and since our product is essentially infinitely complex there is no way we can do exhaustive testing.
This is where you come in. You have your application and your environment. Please, now is the time to begin taking a serious look at whether Java 6 (a.k.a. Mustang) will successfully run your application. Now is the time to notify us of any bugs you deem critical or showstoppers. If you tell us now, we stand a better chance of fixing those bugs than if you tell us in, say, September or October.
OK, here's a couple of regressions:
- JDBC-ODBC time field regression in latest builds causing runtime exceptions
- REGRESSION: JdbcOdbc ResultSet.getString returning text and char(0) padding
And then there's foxtrot that's stopped working as of last week's mustang build (I've mentioned it already in a few places...), looks like a change to the classloader or something similar broke that.
Posted by: chris_e_brown on January 25, 2006 at 01:12 PM
Another one is NPE in MetalInternalFrameUI that has been kind of left behind and is affecting at least one 3rd party look and feel.
Posted by: kirillcool on January 25, 2006 at 11:18 PM
Is that a joke?? I just think the whole blog is more or less ironic, since years I see this calls for "pleas tell use regressions" and every release the bug list gets longer and longer. Since I use java I have at least 20 bugs reported, I think 15 have been accepted some marked as not-a-bug because of my fault and some were ignored. I never counted how many have been fixed but I am sure more than 10 are still open. So why do you call for reporting more bugs if you can't even handle the amounts of bugs already in the database??
The most annoying thing in my eyes is that mustang is more or less a maintenance release, and some areas are quite broken again :(
Posted by: linuxhippy on January 26, 2006 at 08:47 AM
Chris, is the foxtrot issue already submitted to Sun? And thank you for mentioning those, however I'll note in the case of a JDBC issue, that there's third party code involved and the bug may be there (not that I'm saying it is, as I'm not involved with evaluating those bugs).
And, no, this is not a joke. We do want to know about regressions you see. If Mustang is, as you claim, "quite broken again" then we need to know about this. Honestly.
Posted by: robogeek on January 26, 2006 at 09:26 AM
When using tiny look and feel:
swing.aatext=true property doesn't effect menu text. All other text will be anti-aliased, but not menu text. This works fine with 1.5 but doesn't work 1.6. I'm using b68 on Linux. The same issue exists on Windows with b67 with both tiny L&F; and the system (Windows XP) L&F.; I'll test again with b68 on Windows, but I'm expecting the same problem to exist. -Bryan
Posted by: prime21 on January 26, 2006 at 11:46 AM
@Bryan: The problem is in the Mustangs implementation of the
#drawXX methods in
BasicGraphicsUtils and how Mustang enables/handles AA desktop text hints. You can find more information about this problem here:
Menu item anti-aliasing broken on Java 6. You can find also there a possible workaround for this problem. -Andrej
Posted by: golovnin on January 27, 2006 at 12:04 AM
@Bryan: Sorry have forgotten to say, the
swing.aatex property does not more exist in Mustang. AA will be enabled when you use AA in your desktop environment (e.g. when you activate ClearType on Windows). -Andrej
Posted by: golovnin on January 27, 2006 at 12:09 AM
About the Foxtrot issue, I haven't yet submitted it to Sun, because I'm not the owner of the API and as such don't know it well enough in terms of its implementation, and because I posted it on the java.net and javadesktop.org forums, as well as the Foxtrot mailing list (click these links to jump to the relevant threads).
About the JDBC issue, it's with the Sun JDBC-ODBC driver, so it does look like a Sun issue (it worked before, doesn't work any more), not any third-party driver.
As for the "joke", I didn't post that comment... although I do really hope that regression testing and removing bugs in the bug database is a really high priority (I prefer that the existing APIs work before any new APIs are added...).
Posted by: chris_e_brown on January 27, 2006 at 01:14 AM
I appreciate the feedback. However perhaps I wasn't clear enough in the blog posting. We have a public bug entry form, and you can declare when submitting the bug that you believe it is a regression or not. We do take regressions seriously, and part of our criteria before we go to FCS is that regressions are fixed. Known regressions that is.
That's why I said this is not a joke. We want to know about regressions, so we know what to fix.
Of course there is always a question around what constitutes a regression. There are plenty of edge cases where it's hard to make the call.
BTW, watch my blog next week -- if all goes well I'll have an interesting announcement to make then.
Posted by: robogeek on January 27, 2006 at 02:24 PM