On 02/03/2018 02:54 AM, Michał Górny wrote: > W dniu sob, 03.02.2018 o godzinie 00∶57 -0800, użytkownik Zac Medico > napisał: >> On 02/03/2018 12:26 AM, Michał Górny wrote: >>> W dniu pią, 02.02.2018 o godzinie 18∶23 -0800, użytkownik Zac Medico >>> napisał: >>>> There's been a lot of pushback involving the --dynamic-deps=n default. >>> >>> Most of it pushed by a single rogue developer. >>> >>>> What we really need is a tool to apply dependency changes in-place, >>>> without the need for a rebuild. >>> >>> So what is the plan for the 'tool'? Who is working on it? How will it >>> magically work? >> >> The same way as --dynamic-deps=y, but persistently. > > In that case, it's a horrible idea and in violation of the PMS. > Adding a little persistence to the idea that was horribly wrong > in the first place won't suddenly make it right.
I won't bother to defend --dynamic-deps, but I will note that many
people seem to have been happy with if for many years.
>>>> Reverts: 2905e1c2c28d ("Disable dynamic-deps by default")
>>>> Bug: https://bugs.gentoo.org/646458
>>>
>>> This is a horrible idea, and I find it outrageous that you submit it for
>>> review in the middle of the night an then merge yourself 40 minutes
>>> later, and make into a release.
>>
>> After two weeks of listing to people complain about useless rebuilds, I
>> couldn't take it anymore.
>
> Why didn't you commit the news item?
It didn't commit it due to a fair amount of resistance seen in replies.
I'm promoting a 2 phase strategy as follows:
1) Roll out a USE="gentoo-dev" flag that triggers the default
--dynamic-deps=n setting.
2) After gentoo dev's have shown a willingness and ability to do the
revbumps, it will be safe to roll out the default --dynamic-deps=n
setting to regular users.
> It's not surprise they're
> complaining if they don't get the right information, and the next thing
> they see is wall of useless text from --changed-deps-report.
Yes, I've disabled it as you've suggested. When the list exceeds 10
items or so, it will be better to write the full list to a file, and
truncate the terminal output.
>>> The main result we're going to see now is a lot of users hit with
>>> a lot of confusing messages. Yes, you've already made this a mess with
>>> that --changed-deps-report but this is only getting worse. Now nobody
>>> will be able to figure out what way things are supposed to work,
>>> and if they're forced to rebuild everything.
>>
>> Those who are willing to accept the extra rebuilds will do them, and the
>> others will use --dynamic-deps=y until something better comes along.
>>
>>> I'm pretty sure some developers will use this as an excuse to stop
>>> revbumping stuff, and in a few days we're be back at square one.
>>
>> Maybe the QA team can do the revbumps? If devs get enough bug mail
>> involving QA revbumps, maybe they'll eventually cooperate?
>
> Maybe the QA team should just do the work for all the developers?
Maybe not. I'm hoping that my suggested 2 phase strategy will protect
users while we sort out the developer issues.
--
Thanks,
Zac
signature.asc
Description: OpenPGP digital signature
