Re: activities save and restore

2012-12-20 Thread Aaron J. Seigo
On Wednesday, December 19, 2012 23:40:03 Sebastian Kügler wrote:
  and indeed that was it (a QMetaObject::invokeMethod in Start.cpp hidden
  behind  a kDebug())
  fixed it in KDE/4.10 and master, so the desktop should be fine as well

 You rock, Marco!

yep; truly great big bunches of cookies for that one.

also, as this is the second time this has happened,  perhaps you (marco) could
send a note to the author of the commit reminding them that kDebug() does not
get executed in such scenarios? :)

--
Aaron J. Seigo

signature.asc
Description: This is a digitally signed message part.
___
Active mailing list
Active@kde.org
https://mail.kde.org/mailman/listinfo/active


Re: activities save and restore

2012-12-20 Thread Marco Martin
On Thursday 20 December 2012, Aaron J. Seigo wrote:
 On Wednesday, December 19, 2012 23:40:03 Sebastian Kügler wrote:
   and indeed that was it (a QMetaObject::invokeMethod in Start.cpp hidden
   behind  a kDebug())
   fixed it in KDE/4.10 and master, so the desktop should be fine as well
  
  You rock, Marco!
 
 yep; truly great big bunches of cookies for that one.
 
 also, as this is the second time this has happened,  perhaps you (marco)
 could send a note to the author of the commit reminding them that kDebug()
 does not get executed in such scenarios? :)

apparently is not enough, it continues to not work, so it was a combination of 
two problems.

i'll start to bisect again :/

-- 
Marco Martin
___
Active mailing list
Active@kde.org
https://mail.kde.org/mailman/listinfo/active


Re: activities save and restore

2012-12-20 Thread Marco Martin
On Thursday 20 December 2012, Marco Martin wrote:
 On Thursday 20 December 2012, Aaron J. Seigo wrote:
  On Wednesday, December 19, 2012 23:40:03 Sebastian Kügler wrote:
and indeed that was it (a QMetaObject::invokeMethod in Start.cpp
hidden behind  a kDebug())
fixed it in KDE/4.10 and master, so the desktop should be fine as
well
   
   You rock, Marco!
  
  yep; truly great big bunches of cookies for that one.
  
  also, as this is the second time this has happened,  perhaps you (marco)
  could send a note to the author of the commit reminding them that
  kDebug() does not get executed in such scenarios? :)
 
 apparently is not enough, it continues to not work, so it was a combination
 of two problems.
 
 i'll start to bisect again :/

update:
seems to have something to do with the declarative ui for the activity 
encryption.

that ui is pluggable and depending from the usual platform profile from 
kdeclarative it loads a qwidget based one or a qml based one.

if the qwidget based one is loaded, activities are restored correctly.

It seems to make not much sense, because that ui shouldn't have anything to do 
with it.

(to me even raises another question about if we still want to keep the 
activity encryption since isa basically unmaintained, not used and can't work 
reliably due how nepomuk works)

Cheers,
Marco Martin
___
Active mailing list
Active@kde.org
https://mail.kde.org/mailman/listinfo/active


Re: activities save and restore

2012-12-19 Thread Sebastian Kügler
On Tuesday, December 18, 2012 20:19:05 Marco Martin wrote:
 On Tue, Dec 18, 2012 at 7:30 PM, Marco Martin notm...@gmail.com wrote:
 Hi all,
 I'm debugging a problem in active, where activities are not correctly
 restored at boot. then i remembered one thing: also with plasma on desktop
 some weeks ago i had the same problem, of disappearing activities if
 someone is running from master, did this issue happens too? can you see if
 there are empty files (0 length) in
 ~/.kde4/share/apps/plasma-desktop/activities/
 
 I can confirm it's a kactivitymanagerd issue, reverting it with an old rpm
 from active3 fixes the issue completely.
 
 so now the regression has to be found (quickly) since is definitely
 something that may affect the KDE 4.10 release too

I haven't seen this issue on my system. Do you have a reliable way to 
reproduce it (I take it you don't)? If I can reproduce it, I could do the 
bisecting.
-- 
sebas

http://www.kde.org | http://vizZzion.org | GPG Key ID: 9119 0EF9
___
Active mailing list
Active@kde.org
https://mail.kde.org/mailman/listinfo/active


Re: activities save and restore

2012-12-19 Thread Aaron J. Seigo
On Tuesday, December 18, 2012 19:30:04 Marco Martin wrote:
 what happens is that at stop containments are correctly stored (in the
 above directory, one per file)
 then sometimes, it fails to restore the containment leading to an empty
 one, sometimes no containment at all.

if kamd reports the wrong activity information, it is possible that libplasma 
does not load the activity information from the file - empty activity - save 
out - nothing in the file?

-- 
Aaron J. Seigo

signature.asc
Description: This is a digitally signed message part.
___
Active mailing list
Active@kde.org
https://mail.kde.org/mailman/listinfo/active


Re: activities save and restore

2012-12-19 Thread Marco Martin
On Wednesday 19 December 2012, Aaron J. Seigo wrote:
 On Tuesday, December 18, 2012 19:30:04 Marco Martin wrote:
  what happens is that at stop containments are correctly stored (in the
  above directory, one per file)
  then sometimes, it fails to restore the containment leading to an empty
  one, sometimes no containment at all.
 
 if kamd reports the wrong activity information, it is possible that
 libplasma does not load the activity information from the file - empty
 activity - save out - nothing in the file?

i got it building on a virtualbox instance in mer...
and it gets broken only when i do a release build.
that's why we are not seeing this, we are always running a debug build

that means the problem is in some code conditionally built only on debug 
builds
(i remember once a bug in kactivitymanager was an operation done inside a 
kDebug()  statement, it may be that again)

Cheers,
Marco Martin
___
Active mailing list
Active@kde.org
https://mail.kde.org/mailman/listinfo/active


Re: activities save and restore

2012-12-19 Thread Marco Martin
On Wednesday 19 December 2012, Marco Martin wrote:
 On Wednesday 19 December 2012, Aaron J. Seigo wrote:
  On Tuesday, December 18, 2012 19:30:04 Marco Martin wrote:
   what happens is that at stop containments are correctly stored (in the
   above directory, one per file)
   then sometimes, it fails to restore the containment leading to an empty
   one, sometimes no containment at all.
  
  if kamd reports the wrong activity information, it is possible that
  libplasma does not load the activity information from the file - empty
  activity - save out - nothing in the file?
 
 i got it building on a virtualbox instance in mer...
 and it gets broken only when i do a release build.
 that's why we are not seeing this, we are always running a debug build
 
 that means the problem is in some code conditionally built only on debug
 builds
 (i remember once a bug in kactivitymanager was an operation done inside a
 kDebug()  statement, it may be that again)

and indeed that was it (a QMetaObject::invokeMethod in Start.cpp hidden behind 
a kDebug())
fixed it in KDE/4.10 and master, so the desktop should be fine as well

Cheers,
Marco Martin
___
Active mailing list
Active@kde.org
https://mail.kde.org/mailman/listinfo/active


Re: activities save and restore

2012-12-19 Thread Sebastian Kügler
On Wednesday, December 19, 2012 22:37:06 Marco Martin wrote:
 On Wednesday 19 December 2012, Marco Martin wrote:
  On Wednesday 19 December 2012, Aaron J. Seigo wrote:
   On Tuesday, December 18, 2012 19:30:04 Marco Martin wrote:
what happens is that at stop containments are correctly stored (in the
above directory, one per file)
then sometimes, it fails to restore the containment leading to an
empty
one, sometimes no containment at all.
  
   if kamd reports the wrong activity information, it is possible that
   libplasma does not load the activity information from the file - empty
   activity - save out - nothing in the file?
 
  i got it building on a virtualbox instance in mer...
  and it gets broken only when i do a release build.
  that's why we are not seeing this, we are always running a debug build
 
  that means the problem is in some code conditionally built only on debug
  builds
  (i remember once a bug in kactivitymanager was an operation done inside a
  kDebug()  statement, it may be that again)
 
 and indeed that was it (a QMetaObject::invokeMethod in Start.cpp hidden
 behind  a kDebug())
 fixed it in KDE/4.10 and master, so the desktop should be fine as well

You rock, Marco!
-- 
sebas

http://www.kde.org | http://vizZzion.org | GPG Key ID: 9119 0EF9
___
Active mailing list
Active@kde.org
https://mail.kde.org/mailman/listinfo/active


Re: activities save and restore

2012-12-18 Thread Marco Martin
On Tue, Dec 18, 2012 at 7:30 PM, Marco Martin notm...@gmail.com wrote:

 Hi all,
 I'm debugging a problem in active, where activities are not correctly
 restored at boot.
 then i remembered one thing: also with plasma on desktop some weeks ago i
 had the same problem, of disappearing activities
 if someone is running from master, did this issue happens too? can you see
 if there are empty files (0 length) in
 ~/.kde4/share/apps/plasma-desktop/activities/


I can confirm it's a kactivitymanagerd issue, reverting it with an old rpm
from active3 fixes the issue completely.

so now the regression has to be found (quickly) since is definitely
something that may affect the KDE 4.10 release too


Cheers,
Marco Martin
___
Active mailing list
Active@kde.org
https://mail.kde.org/mailman/listinfo/active