RFR: Section on backporting [v7]

[Joe Darcy]

On 12/2/2021 11:56 PM, Goetz Lindenmaier wrote:
> On Fri, 3 Dec 2021 00:36:43 GMT, Jesper Wilhelmsson <jwilhelm at openjdk.org> wrote:
>
>> Right.. So what is the best solution here then?
> Is having twice the same bugId such a big problem?


Some perspective from the design of JBS: a core usage pattern is that a 
bug is (semantically) used for one push to a line of development (with 
some caveats for SCM purposes). That is, work is pushed under one bug id 
and if further work is needed, follow-up changes are  pushed under a 
different bug-id. Obviously different software projects choose different 
conventions in such matters, including not requiring bug ids at all. 
However, the one-push-per-bug (per line of development) convention is 
long-standing JDK feature work and I think has served us reasonably well.

One benefit of this model is its simplicity for common cases. Given the 
bug id, one can look in the bug database for additional context etc.

The model we were aiming for in JBS was to have the per-release 
information be a tuple such as (release, state in the release, one or 
two other fields, ..). However, that wasn't an option in Jira at the 
time so we approximated that model using separate full bugs of a 
distinct backport issue type. The intention was to have the manipulation 
of backport issues be handled primarily by tooling, such as the Skara 
bots and their logical predecessors on the earlier systems.

There are cases where a bug id can be used repeatedly in a repo (e.g. 
fix is pushed in 19 and later backported to 18; when 18 is next synced 
into 19, there will be two changesets with the same bug id). However, 
those cases should be "the same" fix with a different SCM context rather 
than iterations of a fix.

HTH,

-Joe



> The bugid anyways does not refer to the JBS issue connected to the push.
> And these backouts are very rare.
>
> I would propose to remove the backport link and then just do a new backport.  This is the most simple approach for the developers working on the backports. Would this work with skara? Would the repeated bugId break any tooling?
>
> -------------
>
> PR: https://git.openjdk.java.net/guide/pull/66