Where bug fixes land ... Re: OpenJDK 6 and 6u10 features

[David Herron @ Sun]

Andrew John Hughes wrote:
> Yes, I can see the point of porting to 7 instead of 6 because there
> seems to be far more development effort being deployed by Sun on this
> codebase.  From the perspective of shipping a Free JDK now, this
> doesn't make sense but it does from the perspective of developing new
> code.  However, outside participation is still limited on this because
> there is no clearly defined open specification for what will be part
> of JDK 7 yet, and so no clear way for external participation in
> OpenJDK7 more than simple bug fixing.  Such bug fixes are more likely
> to occur on OpenJDK6 because this is what people will tend to run,
> given its stability.
>
> I'm aware of the differences between the two from working with the
> IcedTea versions of both.  Clearly, applying the patches isn't zero
> effort but it is trivial compared to searching out bug reports and
> cleaning up licensing, which is what was discussed in your blog and
> which no-one outside Sun can do.


Andrew, there is also a consideration that (I think) Joe did not 
mention.  OpenJDK6 is an offshoot from the main development trunk.  To 
simplify the long term story it's best that OpenJDK6 be treated as the 
temporary transition tool that it is.  7 is the next step in the normal 
pattern.

There's this picture ... 
http://blogs.sun.com/darcy/entry/openjdk_6_genealogy

The issue I see here in what you've written is the question of where bug 
fixes land.

I do agree that a bug will be observed in OpenJDK<n> while development 
will occur in OpenJDK<n+1>.  That's because <n> is the deployed release 
while <n+1> is not meant to be the deployed production use release.

Our normal practice is to put bugfixes into <n+1>, let them soak awhile 
to get some testing, and if warranted backport to <n> or earlier 
releases (<n-1>, <n-2>, ...).  Of course there are instances where it's 
best to follow a different pattern, this is simply the normal pattern.  
Security fixes for instance follow a pattern of being applied to all 
releases simultaneously.

What are you suggesting as the pattern for applying bugfixes?

- David