[Bug 126073] drkonqi blocks logout if some application crashes

[Jekyll Wu]

https://bugs.kde.org/show_bug.cgi?id=126073

--- Comment #42 from Jekyll Wu <adaptee at gmail.com> ---
(In reply to comment #41)
> i am not quite sure if i understand the description correctly -- but the
> issue is that drkonqi _blocks_ logout.

No, quite the opposite. Read the review quest and the commit link,  then you
will see that it was "startkde" that decided to wait fo drkonqi forever. There
is no way for drkonqi itself to block the shutdown process.  "startkde" is just
a shell script, not hard to read.


> while a crash of any application upon logout is nasty and possibly deserves
> research, the core issue is the failing logout/shutdown -- thus, a timeout
> of 15 _minutes_ waiting for any response to drkonqi is not a solution.

Not a real solution, but much better than the previous "waiting forever"
situation. 

> 
> imo drkonqi should do nothing at all:
> - usually there are no debug libs installed, hence there will be no usable
> stacktrace anyway

No, imcomplete backtrace without debug symbols are still much more useful than
no bakctrace at all.  If drkonqi just does nothing (not even showing up), then
users  :

  *  maybe not even realize something has crashed during shutdown thus never
report it . 
  *  cant not provide any useful infomation (backtrace, even if incomplete)
except "something crashed during shutdown".

That make the crash less likely to be spotted, reported, investigated and
fixed. 


> - user wants to logout/shutdown -- that's most likely not the time for
> prolonged waiting or activities

If he/she notice that but doesn't have time/patience at that moment, just close
drkonqi with one click.  No extra time wasted for he/she. If he/she does has
time/patience at that time, then why not give he/she the chance of
reporting/saving the crash information ? 


> instead, a check for a usable stacktrace should be performed (or when
> possible a check for debugging facilities) -- if unavailable skip the whole
> drkonqi stuff. 

Again, that kind of smart checking should *only* happen in this special
shutdown case, but not in average cases . And detecting that shutdown case (I
think) is tricky and unnecessary work for drkonqi (see the review request).  
The better solution is drkonqi provides more dbus methods and startkde uses
them in a smarter way than now. But that might take long time to implement or
never happen.

-- 
You are receiving this mail because:
You are the assignee for the bug.