Re: [gentoo-portage-dev] [PATCH] emerge: add --changed-deps-report option (bug 645780)

[Zac Medico]

On 01/28/2018 12:16 PM, Michał Górny wrote:
> W dniu nie, 28.01.2018 o godzinie 10∶09 -0800, użytkownik Zac Medico
> napisał:
>> On 01/28/2018 09:35 AM, Alec Warner wrote:
>>>
>>>
>>> On Sun, Jan 28, 2018 at 9:51 AM, Zac Medico <zmedico@g.o
>>> <mailto:zmedico@g.o>> wrote:
>>>
>>>     The --dynamic-deps=n default causes confusion for users that are
>>>     accustomed to dynamic deps, therefore add a --changed-deps-report
>>>     option that is enabled by default (if --usepkgonly is not enabled).
>>>
>>>     The --quiet option will suppress the report if none of the packages
>>>     having changed dependencies are in the dependency graph, since they
>>>     are harmless in that case. If any of these packages *are* in the
>>>     dependency graph, then --quiet suppresses the big NOTE section of
>>>     the report, but the HINT section is still displayed since it might
>>>     help users resolve problems that are solved by --changed-deps.
>>>
>>>     Example output is as follows:
>>>
>>>     !!! Detected ebuild dependency change(s) without revision bump:
>>>
>>>         net-misc/openssh-7.5_p1-r3::gentoo
>>>         sys-fs/udisks-2.7.5::gentoo
>>>
>>>     NOTE: For the ebuild(s) listed above, a new revision may be
>>>     warranted if there
>>>           has been a dependency change with significant consequences.
>>>     Refer to the
>>>           following page of the Gentoo Development Guide for examples of
>>>           circumstances that may qualify:
>>>
>>>
>>>     https://devmanual.gentoo.org/general-concepts/ebuild-revisions/
>>>     <https://devmanual.gentoo.org/general-concepts/ebuild-revisions/>
>>>
>>>           If circumstances qualify, please report a bug which specifies
>>>     the current
>>>           version of the ebuild listed above. Changes to ebuilds from
>>>     the 'gentoo'
>>>           repository (ending with '::gentoo') can be browsed in GitWeb:
>>>
>>>               https://gitweb.gentoo.org/repo/gentoo.git/
>>>     <https://gitweb.gentoo.org/repo/gentoo.git/>
>>>
>>>           Use Gentoo's Bugzilla to report bugs only for the 'gentoo'
>>>     repository:
>>>
>>>               https://bugs.gentoo.org/
>>>
>>>           In order to suppress reports about dependency changes, add
>>>           --changed-deps-report=n to the EMERGE_DEFAULT_OPTS variable in
>>>           '/etc/portage/make.conf'.
>>>
>>>     HINT: In order to avoid problems involving changed dependencies, use the
>>>           --changed-deps option to automatically trigger rebuilds when
>>>     changed
>>>           dependencies are detected. Refer to the emerge man page for more
>>>           information about this option.
>>>
>>>     Bug: https://bugs.gentoo.org/645780
>>>
>>>
>>> I can't really support sending this large report to users.
>>>
>>> 1) Its fairly common practice today.
>>> 2) All users will get the report.
>>> 3) A subset of them will file bugs about the report.
>>> 4) Devs make a decision about revbumping vs not; there doesn't seem to
>>> be a way for devs to say "no this change is intentional, stop nagging
>>> users."
>>
>> I think in practice we need to revbump for most changes. If the changes
>> weren't worth propagating, then we wouldn't make them.
>
> It's not about 'being worth propagating', it's about 'being worth
> the rebuild to propagate'.
>
> I've bumped dependency inside all LLVM ebuilds today. The change fixes
> only problem that hits people who:
>
> a. don't use --deep when upgrading,
>
> b. use clang with LTO.
>
> I can't say how many people were hit by it since 5.0.0 was introduced
> but yesterday I've got the first (and only) report so far.
>
> Yes, I could technically revbump and cause people to have to spend most
> of the day rebuilding 1-2 versions of LLVM for change that doesn't make
> any difference to them.

I can see how that would be troublesome for someone who runs ~arch and
syncs once or more per day. For a stable user that syncs once a week or
less often, it's an entirely different story.

> Yes, I could consider the change 'not worth making'. But why shouldn't I
> improve stuff for our users when it doesn't harm to do so?
>
> But with your suggested solution, now we no longer have the choice
> 'revbump or not revbump'. Now we have a choice between forcing people to
> rebuild via revbump or forcing people to get verbose report that most
> likely will result in meaningless bug report and/or rebuild. So I end up
> having a choice between 'force people to rebuild' or 'not fix minor bugs
> at all'.

Instead of suggesting to file a bug report, I think it should link to a
wiki page that we will allow us to easily refine the content over time.
Hopefully that will address everyone's concerns.

My main intent is to provide a safety net for users, such that they'll
be automatically notified that that --changed-deps option is available
when needed to solve problems involving stale dependencies.
-- 
Thanks,
Zac