Should back-ports cc the mailing lists for the area? (was Re: [7u8] Request for approval: 7113017: Use POSIX compliant include file headers in sun/awt/medialib/mlib_types.h)

[Neil Richards]

On Tue, 2012-08-07 at 16:03 +0100, Seán Coffey wrote:
> +1 on Edvard's suggestion then. Giving the original author notification 
> of the proposed backport first will help avoid the problem described. 
> There's also the option to request extra reviews at 7u approval time 
> also if needed (as per 7u ground rules)
> 
> Alan makes a good point around the original jdk8 author not always being 
> the correct one to ask. In that case, should the list of engineers to be 
> notified of backport extend to author & original reviewers ? I conscious 
> of not wanting to spam the various -dev aliases with each backport 
> request that may be occurring with 8 -> 7u.
> 
> regards,
> Sean.
> 
> On 07/08/2012 14:25, Alan Bateman wrote:
> > On 07/08/2012 13:51, Edvard Wendelin wrote:
> >> Hi,
> >>
> >> When I've been working on backports that I'm not the original author 
> >> of, I have always asked the engineer if the fix is suitable to 
> >> backport or not. Instead of having to CC the entire group I'd like to 
> >> suggest the following alternative rule:
> >>
> >> If you are not the original author, make sure both of you are in 
> >> agreement that the fix is suitable for backport before requesting for 
> >> approval on the mailing list.
> >>
> >> I think the most common use case by far is that the same engineer 
> >> takes the fix to 8 and 7u and in that case we shouldn't add more 
> >> processes.
> >>
> > I don't want to waste any time on this but I think the concern is 
> > mostly with cases where the original author doesn't have history in 
> > the specific area where the change is. He/she gets a review from 
> > someone in that area for jdk8 and then gets approval (!= code review) 
> > for the change for jdk7u-dev as that's where they are mostly 
> > interested in. In that case the original reviewers/experts don't 
> > always see it or only see it after it's in jdk7u-dev. I think this is 
> > what Phil's original complaint was about. I think Andrew and David's 
> > concerns are slightly different and covered in the previous thread.
> >
> > -Alan.
> 

I wonder if more use could be made of the "related bugs" field of a java
bug.

As I recall, the author of a fix gets (e-mail) notification when info is
modified in the associated java bug.
This might be the prompt for them to investigate the related bug (if
only to exonerate their code!). 

Having links to unresolved related bugs would also be an indication to
any reviewer that the original code isn't yet ready for backport
approval.

(I suppose I'm assuming that bugs which are linked as "related" in this
field can subsequently be unlinked again, if the relationship is proven
false).

I don't know what search facilities are available inside Oracle, but I
haven't figured out a way of searching for one java bug number in the
comments of another via bugs.sun.com
(The "search" field doesn't like numbers, the "bug id" field just takes
you to the bug number you type in).

So without this linking back to the bug for the original fix (or a
posting to the mailing list that there's a potential problem with it)
there's no sensible externally-observable indication that anything's
gone awry.

(NB: in the case that prompted this conversation, it was the original
author who suggested the backport, and the original change had been the
jdk8 codebase for over 7 months prior to the backport suggestion being
made. There was no indication on the mailing lists or in the change's
java bug that there was any concern over it).

I agree that including the area-specific mailing list in the backport
requests is a useful additional backstop / check, but hopefully better
communication prior to that point will help to both prevent requests for
changes under suspicion from being made in the first place, and also
engage more eyes sooner to work out resolutions for the related bug(s).

Regards,
Neil

-- 
Unless stated above:
IBM email: neil_richards at uk.ibm.com
IBM United Kingdom Limited - Registered in England and Wales with number 741598.
Registered office: PO Box 41, North Harbour, Portsmouth, Hampshire PO6 3AU