[KDE/Mac] DrKonqi placement

[Ian Wadham]

Hi Marko,

Nice to see you again… :-)

On 23/10/2014, at 7:30 AM, Marko Käning wrote:
> I felt urged to forward this to kde-mac...
> 
> But, I guess you are aware of this discussion thread on k-f-d, aren’t you?

No, I don't follow kde-frameworks-devel, but it's good that one of us does…
This topic is of interest on the OSX platform and probably Windows too.  It
should really be discussed on kde-devel or at least kde-core-devel.

> On 22 Oct 2014, at 09:13 , David Faure <faure at kde.org> wrote:
>> On Monday 20 October 2014 10:54:51 Milian Wolff wrote:
>>> On Wednesday 02 April 2014 18:48:24 Aleix Pol wrote:
>>>> Hi,
>>>> I know I'm changing what I say about drkonqi every now and then, but every
>>>> step I take, I realize that the project is bigger and bigger. What I
>>>> wanted
>>>> to do originally was to move it to KCrash, KCrash without drkonqi is much
>>>> less useful.
>>> 
>>> <snip>
>>> 
>>>> - with bugzilla integration (additionally to the previous ones):
>>>> KF5::WebKit
>>> 
>>> Could this maybe be replaced by using QtWebKit directly?
>> 
>> Then it wouldn't use KWallet, KIO, KCookieJar...

BTW, KCookieJar is no longer relevant to Dr Konqi.  Bugzilla on bugs.kde.org
stopped issuing or accepting cookies in July.  That is what all the hoo-ha and
last-minute panic was about regarding https://git.reviewboard.kde.org/r/120431/
 "Fix and future-proof Dr Konqi security methods on Bugzilla"… :-)
(@David: I am the guy who submitted that patch).

KIO comes in as a tail-end I/O method after Dr Konqi sends a remote procedure
call to Bugzilla via the KXmlRpc::Client class.  KXmlRpc::Client could presumably
use another I/O method --- carrier pigeons maybe… :-)  Or perhaps, Dr K could use
another Bugzilla command-protocol, such as REST (but I do not know much about that).

>> Would be annoying to have to enter your password again there, when the KIO 
>> infrastructure would normally take care of it….

KWallet is all you need to get your login ID and password entered in Dr Konqi's
login dialog box.  It even works on Apple OS X now… ;-)

In future versions of Bugzilla, the login ID and password will be required on
every RPC that updates the database.  AFAICT, from scanning the Dr Konqi
code, there will usually be one such call per crash, with the alternatives being
to submit a new bug report or add an attachment to an existing report.

If that is so, the login dialog should be relocated so as to occur immediately
before the database update (if there is one).  That way, we could avoid having
to log in unnecessarily, if the outcome is that the crash has insufficient or
redundant information and Dr Konqi decides not to report it.

Dr K should still say, early in the piece, that a bugs.kde.org account/login could
be needed, so that the end-user is prepared for that before he goes through all
the dialog boxes.

>> Maybe we need to separate the basic crash dialog from the bug reporting 
>> functionality….

It is already separated quite a lot.  The real problem IMHO is that the succession
of dialog boxes and the interaction between them and the underlying database
(bugs.kde.org) has been coded in such a complex way.  Perhaps that is because
the underlying "wizard" paradigm is sequential, whereas the dialog is branching
and has (at least) three outcomes: report a bug, add to an existing bug or report
nothing (insufficient information).  Also, each branch can depend on complex
evaluation of crash information, possible duplicates, etc.

All the best, Ian W.

>> David Faure, faure at kde.org, http://www.davidfaure.fr
>> Working on KDE Frameworks 5