[QGIS-Developer] backporting nightmare

[Matthias Kuhn]

On 3/4/20 8:04 AM, Nyall Dawson wrote:
> On Wed, 4 Mar 2020 at 17:01, Matthias Kuhn <matthias at opengis.ch> wrote:
>> On 3/4/20 7:48 AM, Denis Rouzaud wrote:
>>
>>
>>
>> Le mer. 4 mars 2020 à 02:56, Nyall Dawson <nyall.dawson at gmail.com> a écrit :
>>> On Wed, 4 Mar 2020 at 10:48, William Kyngesburye <woklist at kyngchaos.com> wrote:
>>>
>>>> If it's just a buggy backport bot, then that's bad timing on my part, after all my dwindling spare time it took me to get caught up on my Mac build, I get this.
>>> It is. The bot got seriously crippled by an upstream change, and now
>>> it's a coin toss on the results you'll get.
>>
>> I plan to give another approach a try during the HF.
>> The idea would be to use a CLI tool (https://github.com/sqren/backport or https://github.com/chainer/backport) which would be run a cron workflow.
>> Using a cron will avoid the issues of PR from forks (the workflow user has rights to create branches).
>> Hopefully these tools work better than the current bot.
>> The only drawback I see up to now is that you'll have to wait for the cron to start (30min?).
>>
>>
>> It looks like the bot fails on cherry-picking merge commits and would (likely) do fine if a `-m` option was provided - at least that's what's missing from the commands in the comment.
>>
>> Maybe it's as easy as adding this?
> it still depends on "merges" instead of rebases... and I'm not a fan
> of merge commits in general. They really make the git history messy
> and difficult to read...
>
> Nyall

Then it was too good to be true. I miss the old backport bot.

I am generally a fan of merges, mostly because they provide a missing 
link from "a set of commits" to a "consolidated feature" aka pull 
request with discussion. We don't have that with rebase merges I think?

Matthias