Wireshark-dev: Re: [Wireshark-dev] What about backporting fixes to older releases with the new
From: Bálint Réczey <balint@xxxxxxxxxxxxxxx>
Date: Wed, 2 Apr 2014 13:25:58 +0200
Hi Gerald & All,

2014-04-01 2:41 GMT+02:00 Pascal Quantin <pascal.quantin@xxxxxxxxx>:
> 2014-03-31 23:17 GMT+02:00 Gerald Combs <gerald@xxxxxxxxxxxxx>:
>
>> On 3/27/14 10:13 PM, Anders Broman wrote:
>> > Hi,
>> > How do we handle backports in the new work flow with git? The submitter
>> > of a patch could help
>> > by submitting the backport once the patch has been accepted. But what do
>> > we do in the case
>> > when this isn't happening? The core developer accepting the patch might
>> > not have the time/don't want
>> > the extra work of making a backport.
>>
>> Prior to the Git migration the Roadmap page was effectively a dumping
>> ground for merge conflicts. I'd end up processing the queue a day or so
>> before each release which didn't leave much time for testing or
>> validation. I would very much like to avoid going back to that.
>>
>> If there aren't any merge conflicts you can cherry-pick a change using
>> several methods:
>>
>> - "git cherry-pick"
>> - The "Cherry Pick To" button in Gerrit's web interface
>> - The "gerrit-cherry-pick" script:
>>    https://code.wireshark.org/review/Documentation/cmd-cherry-pick.html
>> - git-review's "-x" flag
>>
>> For each cherry-pick the release notes need to be updated with any bug
>> fixes, protocol updates and (if needed) an advisory. This can be done by
>> amending or with a separate commit. I don't think this is documented
>> anywhere but I can add instructions to the wiki and/or the Developer's
>> Guide.
>>
>> If there are merge conflicts someone needs to decide if the backport is
>> worth the effort of resolving the conflict. Again, I would prefer that
>> this happens as early as possible.
>
>
> Hi Gerald,
>
> what about using the old wiki page to list the bigfixes candidates for
> backport but not done yet? This could help (temporarily) people not
> confident yet with the new backport procedure (even if git cherry pick
> command makes it much less error prone than subversion).
> BTW sorry for not updating the release notes with my backports (it was not
> documented afaik).
I think this partial resurrection of the old wiki page would be a wise
idea. It also listed the expected date of next release which would be
still useful.

IMO it would be better to update the documentation when right before
the release instead of with every commit because it would make
cherry-picking changes easier (like master -> master-1.10 ->
master-1.8, etc).

Cheers,
Balint