Re: [gnome-settings-daemon] datetime: Remove datetime D-Bus mechanism

[Bastien Nocera <hadess hadess net>]

Em Sat, 2012-01-28 às 11:52 -0500, Colin Walters escreveu:
> On Sat, 2012-01-28 at 15:31 +0000, Bastien Nocera wrote:
> 
> > Ubuntu has apparently taken steps to fix the problem, Debian can
> > probably use the same code, and OpenSUSE can use the RPM I made and
> > linked to earlier in the thread as a basis to fixing the problem.
> > 
> > You call me disrespectful and qualify my attitude as cavalier. It's
> > anything but. There are many ways to solve this problem, and plenty that
> > don't involve e-mails to me or reverting commits I made. I'd like those
> > to be contemplated.
> 
> It is about email - just doing "git log plugins/datetime/" one can see
> the people who had contributed to that code, and as far as I can see,
> they were not all notified that it was just going to be deleted.  Not
> all of these people watch bugzilla for g-s-d consistently.
> 
> For example both Michael Terry and Vincent Untz had contributed to the
> code, but they seemed surprised.  And they contribute to GNOME, and for
> them it's a regression.

I don't think Vincent was surprised, or he really shouldn't have been:
https://bugzilla.gnome.org/show_bug.cgi?id=654970#c2
https://bugzilla.gnome.org/show_bug.cgi?id=654970#c3

Michael didn't comment on this issue in this thread. Sebastien is CC:ed
on gnome-settings-daemon bugs.

> That's the point here - we need to minimize regressions.  No one wants
> to be part of a project where other people come along and either
> unintentionally or intentionally remove/break code that you wrote that
> worked.
> 
> And this is just shuffling around where the date/time code lives.  It's
> hard to describe it as a new, compelling addition to Free Software.

It's also about removing 3k lines of code that lived in the wrong place
in the stack for all these years (it was in gnome-panel before). And
about fixing the date & time panel to stop using synchronous APIs.

> Anyways I don't want to drag this out more than it already is - if the
> affected people are OK implementing a DBus service, then that works.
> But if they feel e.g. that it's easier for them to keep the mechanism
> code around and have control-center talk to it, we should at least
> consider that perspective, because that code worked before, and could
> continue working.

It worked for 4 distros. It can work for those 4 again with little work.

/Bastien, looking forward to removing support for fallback mode