I just discovered that keyboard shortcuts that are meant to affect the active window (such as using ALT+F4 to close) work only on windows with a bold task bar button (like Konversation in the screenshot), however those same keyboard shortcuts will *not* work on an active window which does not have a bold task bar button (like KEdit in the screenshot).

I'm pretty sure I fixed this issue last night:
https://git.trinitydesktop.org/cgit/tdebase/commit/?id=17b142dd646ce4053f4548c0ff7b189531219c32

What I still don't understand is why it only seemed to first appear between RC1 and RC2--I have gone over the twin changes very carefully and nothing stands out.

(In reply to Timothy Pearson from comment #2)
> I'm pretty sure I fixed this issue last night:
> https://git.trinitydesktop.org/cgit/tdebase/commit/
> ?id=17b142dd646ce4053f4548c0ff7b189531219c32
> 
> What I still don't understand is why it only seemed to first appear between
> RC1 and RC2--I have gone over the twin changes very carefully and nothing
> stands out.

It is possible that it was caused by the commit 84d7db0?
https://git.trinitydesktop.org/cgit/tdebase/diff/twin/activation.cpp?id=84d7db0

Note, this part was reverted by commit a7c3b934 because problems with twin crashes.
https://git.trinitydesktop.org/cgit/tdebase/commit/?id=a7c3b934

(In reply to Slávek Banko from comment #3)
> (In reply to Timothy Pearson from comment #2)
> > I'm pretty sure I fixed this issue last night:
> > https://git.trinitydesktop.org/cgit/tdebase/commit/
> > ?id=17b142dd646ce4053f4548c0ff7b189531219c32
> > 
> > What I still don't understand is why it only seemed to first appear between
> > RC1 and RC2--I have gone over the twin changes very carefully and nothing
> > stands out.
> 
> It is possible that it was caused by the commit 84d7db0?
> https://git.trinitydesktop.org/cgit/tdebase/diff/twin/activation.
> cpp?id=84d7db0
> 
> Note, this part was reverted by commit a7c3b934 because problems with twin
> crashes.
> https://git.trinitydesktop.org/cgit/tdebase/commit/?id=a7c3b934

That was my original guess, but then the problems persisted even with a7c3b934 applied.  In any case 17b142dd cured the problems on my test machine (along with the original subject of this bug report); I just need confirmation of this from others before closing out this report.

Thanks!

(In reply to Timothy Pearson from comment #4)
> (In reply to Slávek Banko from comment #3)
> > (In reply to Timothy Pearson from comment #2)
> > > I'm pretty sure I fixed this issue last night:
> > > https://git.trinitydesktop.org/cgit/tdebase/commit/
> > > ?id=17b142dd646ce4053f4548c0ff7b189531219c32
> > > 
> > > What I still don't understand is why it only seemed to first appear between
> > > RC1 and RC2--I have gone over the twin changes very carefully and nothing
> > > stands out.
> > 
> > It is possible that it was caused by the commit 84d7db0?
> > https://git.trinitydesktop.org/cgit/tdebase/diff/twin/activation.
> > cpp?id=84d7db0
> > 
> > Note, this part was reverted by commit a7c3b934 because problems with twin
> > crashes.
> > https://git.trinitydesktop.org/cgit/tdebase/commit/?id=a7c3b934
> 
> That was my original guess, but then the problems persisted even with
> a7c3b934 applied.  In any case 17b142dd cured the problems on my test
> machine (along with the original subject of this bug report); I just need
> confirmation of this from others before closing out this report.
> 
> Thanks!

Sorry, above that should read "(along with the original subject of Bug 2189)" above.

Tim

(In reply to Timothy Pearson from comment #5)
> (In reply to Timothy Pearson from comment #4)
> > (In reply to Slávek Banko from comment #3)
> > > (In reply to Timothy Pearson from comment #2)
> > > > I'm pretty sure I fixed this issue last night:
> > > > https://git.trinitydesktop.org/cgit/tdebase/commit/
> > > > ?id=17b142dd646ce4053f4548c0ff7b189531219c32
> > > > 
> > > > What I still don't understand is why it only seemed to first appear between
> > > > RC1 and RC2--I have gone over the twin changes very carefully and nothing
> > > > stands out.
> > > 
> > > It is possible that it was caused by the commit 84d7db0?
> > > https://git.trinitydesktop.org/cgit/tdebase/diff/twin/activation.
> > > cpp?id=84d7db0
> > > 
> > > Note, this part was reverted by commit a7c3b934 because problems with twin
> > > crashes.
> > > https://git.trinitydesktop.org/cgit/tdebase/commit/?id=a7c3b934
> > 
> > That was my original guess, but then the problems persisted even with
> > a7c3b934 applied.  In any case 17b142dd cured the problems on my test
> > machine (along with the original subject of this bug report); I just need
> > confirmation of this from others before closing out this report.
> > 
> > Thanks!
> 
> Sorry, above that should read "(along with the original subject of Bug
> 2189)" above.
> 
> Tim

I would try to confirm, except that twin is not appearing in today's dist-upgrade. About how long should it take for the changes to hit the nightlies repo?

(In reply to Kristopher from comment #6)
> I would try to confirm, except that twin is not appearing in today's
> dist-upgrade. About how long should it take for the changes to hit the
> nightlies repo?

It's building as I type this, so probably a few hours longer at most.

Tim

(In reply to Timothy Pearson from comment #7)
> (In reply to Kristopher from comment #6)
> > I would try to confirm, except that twin is not appearing in today's
> > dist-upgrade. About how long should it take for the changes to hit the
> > nightlies repo?
> 
> It's building as I type this, so probably a few hours longer at most.
> 
> Tim

This problem is semi-fixed. I am able to properly switch tasks and use keyboard shortcuts, and the task bar buttons update properly upon switching tasks.

Transparency effects still don't update upon switching windows, and newly opened windows are not marked as active on the task bar until I switch away from the window then switch back (though I can interact with the window normally prior to switching).

Sleavek, can you do a GIT bisect from RC1 to current GIT to figure out which commit introduced this defect?  Looks like we'll be shipping RC2 with this bug but we need to find out why it is present to fix it for R14 final.

Thanks!

(In reply to Timothy Pearson from comment #9)
> Sleavek, can you do a GIT bisect from RC1 to current GIT to figure out which
> commit introduced this defect?  Looks like we'll be shipping RC2 with this
> bug but we need to find out why it is present to fix it for R14 final.
> 
> Thanks!

Sorry for fat-fingering your name above.  I can reproduce the bug here so I'm running the bisect (as I likely introduced the problem in the first place it's only fair...)

(In reply to Timothy Pearson from comment #10)
> (In reply to Timothy Pearson from comment #9)
> > Sleavek, can you do a GIT bisect from RC1 to current GIT to figure out which
> > commit introduced this defect?  Looks like we'll be shipping RC2 with this
> > bug but we need to find out why it is present to fix it for R14 final.
> > 
> > Thanks!
> 
> Sorry for fat-fingering your name above.  I can reproduce the bug here so
> I'm running the bisect (as I likely introduced the problem in the first
> place it's only fair...)

I ran a bisect on twin all the way back to GIT hash 90f1f50f (11/28/2013); these are my results...
1.) For the bug where a newly created window cannot be minimized by clicking on its taskbar entry, this was introduced in GIT hash 17b142dd.
2.) For the bug where a previously minimized and restored window cannot be minimized by clicking on its taskbar entry, this was introduced sometime before GIT hash 90f1f50f (bisecting before that hash is much more difficult an time-consuming; it would be easiest to see if that bug existed in 3.5.13.2 and work forward).  This bug was fixed in GIT hash 17b142dd which (irritatingly) also caused the previous bug.

Kristopher, you say task switching worked fine in RC1?  I ask because the issue #2 above existed long before RC1 according to my twin bisect...

Thanks!

Tim

(In reply to Timothy Pearson from comment #11)
<snip>
> Kristopher, you say task switching worked fine in RC1?  I ask because the
> issue #2 above existed long before RC1 according to my twin bisect...
> 
> Thanks!
> 
> Tim

I cannot say for certain since I never actually installed RC1 -- I have been using the nightlies since months before RC1 was announced, and haven't bothered switching to the RC1 repo. However, I *did* do a dist-upgrade on the same day that RC1 was announced, and that's roughly when I noticed this bug appear.

So far as I can see, RC1 packages are for Ubuntu only, a distro I personally won't use (I've had too many issues with it), but if RC1 packages are available for either Debian or CentOS, I can test in a VM to see if this bug is present there.

(In reply to Kristopher from comment #12)
> (In reply to Timothy Pearson from comment #11)
> <snip>
> > Kristopher, you say task switching worked fine in RC1?  I ask because the
> > issue #2 above existed long before RC1 according to my twin bisect...
> > 
> > Thanks!
> > 
> > Tim
> 
> I cannot say for certain since I never actually installed RC1 -- I have been
> using the nightlies since months before RC1 was announced, and haven't
> bothered switching to the RC1 repo. However, I *did* do a dist-upgrade on
> the same day that RC1 was announced, and that's roughly when I noticed this
> bug appear.

Can you narrow the bug appearance down any further than around RC1 release?

> So far as I can see, RC1 packages are for Ubuntu only, a distro I personally
> won't use (I've had too many issues with it), but if RC1 packages are
> available for either Debian or CentOS, I can test in a VM to see if this bug
> is present there.

RC1/RC2 have Debian Wheezy packages available.  Final will have Debian Squeeze and Jessie too.

Tim

(In reply to Timothy Pearson from comment #13)
> (In reply to Kristopher from comment #12)
> > (In reply to Timothy Pearson from comment #11)
> > <snip>
> > > Kristopher, you say task switching worked fine in RC1?  I ask because the
> > > issue #2 above existed long before RC1 according to my twin bisect...
> > > 
> > > Thanks!
> > > 
> > > Tim
> > 
> > I cannot say for certain since I never actually installed RC1 -- I have been
> > using the nightlies since months before RC1 was announced, and haven't
> > bothered switching to the RC1 repo. However, I *did* do a dist-upgrade on
> > the same day that RC1 was announced, and that's roughly when I noticed this
> > bug appear.
> 
> Can you narrow the bug appearance down any further than around RC1 release?

As mentioned in my comment, the same day as the announcement. That's as close as I can get, and I cannot know how "in sync" the nightlies were with RC1 on that day.

> > So far as I can see, RC1 packages are for Ubuntu only, a distro I personally
> > won't use (I've had too many issues with it), but if RC1 packages are
> > available for either Debian or CentOS, I can test in a VM to see if this bug
> > is present there.
> 
> RC1/RC2 have Debian Wheezy packages available.  Final will have Debian
> Squeeze and Jessie too.
> 
> Tim

Only Ubuntu repos are listed in the wiki article linked to from the news article, and searching the TDE wiki for Debian turns up instructions only for stable and nightlies.

(In reply to Kristopher from comment #14)
> As mentioned in my comment, the same day as the announcement. That's as
> close as I can get, and I cannot know how "in sync" the nightlies were with
> RC1 on that day.

OK, just making sure. :-)  The best I can do is to mention that the problem does not seem to exist as of builds using this commit (08/07/2014):
https://git.trinitydesktop.org/cgit/tdebase/commit/?id=d5d78f73c41268a45e63c5dff8f84fe4576748d1

> Only Ubuntu repos are listed in the wiki article linked to from the news
> article, and searching the TDE wiki for Debian turns up instructions only
> for stable and nightlies.

Follow the Ubuntu instructions but substitute "wheezy" for "trusty".

(In reply to Timothy Pearson from comment #15)
> (In reply to Kristopher from comment #14)
> > As mentioned in my comment, the same day as the announcement. That's as
> > close as I can get, and I cannot know how "in sync" the nightlies were with
> > RC1 on that day.
> 
> OK, just making sure. :-)  The best I can do is to mention that the problem
> does not seem to exist as of builds using this commit (08/07/2014):
> https://git.trinitydesktop.org/cgit/tdebase/commit/
> ?id=d5d78f73c41268a45e63c5dff8f84fe4576748d1

Found another system that was updated later on (but still before RC1); seems to be OK as of these hashes:
tdelibs: https://git.trinitydesktop.org/cgit/tdelibs/commit/?id=8814b54e2bc964d6000c156e64af2b125dcc816a
tdebase: https://git.trinitydesktop.org/cgit/tdebase/commit/?id=4866380d2111a66b16a251ab6540affcc135992e

Also, the bug is not isolated to twin--reverting and rebuilding only twin does *not* repair the defect on my test system.

I did some further testing with clean profiles; it seems that the bad commits changed something in one or more configuration files that led to the issues originally described in this report.  Testing with a clean profile did not show any issues with commit a7c3b934, but confirmed problems introduced in GIT hash 17b142dd.

Something is overall just not right in the twin focus handling code though!

(In reply to Timothy Pearson from comment #17)
> I did some further testing with clean profiles; it seems that the bad
> commits changed something in one or more configuration files that led to the
> issues originally described in this report.  Testing with a clean profile
> did not show any issues with commit a7c3b934, but confirmed problems
> introduced in GIT hash 17b142dd.
> 
> Something is overall just not right in the twin focus handling code though!

Which window decorations are you using?  It seems the issues stem from the Crystal window decorations on my test box.

Thanks!

(In reply to Timothy Pearson from comment #18)
> (In reply to Timothy Pearson from comment #17)
> > I did some further testing with clean profiles; it seems that the bad
> > commits changed something in one or more configuration files that led to the
> > issues originally described in this report.  Testing with a clean profile
> > did not show any issues with commit a7c3b934, but confirmed problems
> > introduced in GIT hash 17b142dd.
> > 
> > Something is overall just not right in the twin focus handling code though!
> 
> Which window decorations are you using?  It seems the issues stem from the
> Crystal window decorations on my test box.
> 
> Thanks!

I backed out 17b142dd and put in a gentler solution to Bug 2189 instead.  This will be part of R14 RC2.

I still need to know which window decorations you are using to debug the original report further.

Thanks!

(In reply to Timothy Pearson from comment #19)
> (In reply to Timothy Pearson from comment #18)
> > (In reply to Timothy Pearson from comment #17)
> > > I did some further testing with clean profiles; it seems that the bad
> > > commits changed something in one or more configuration files that led to the
> > > issues originally described in this report.  Testing with a clean profile
> > > did not show any issues with commit a7c3b934, but confirmed problems
> > > introduced in GIT hash 17b142dd.
> > > 
> > > Something is overall just not right in the twin focus handling code though!
> > 
> > Which window decorations are you using?  It seems the issues stem from the
> > Crystal window decorations on my test box.
> > 
> > Thanks!
> 
> I backed out 17b142dd and put in a gentler solution to Bug 2189 instead. 
> This will be part of R14 RC2.
> 
> I still need to know which window decorations you are using to debug the
> original report further.
> 
> Thanks!

I am using Crystal with transparency enabled in Crystal's background settings. I have also enabled translucency under Control Center -> Window Behavior -> Translucency. If you need more specific details, let me know (or just point me to the config files to attach).

If you still need me to test using RC1, I should be able to get to it later today.

(In reply to Kristopher from comment #20)
> I am using Crystal with transparency enabled in Crystal's background
> settings. I have also enabled translucency under Control Center -> Window
> Behavior -> Translucency. If you need more specific details, let me know (or
> just point me to the config files to attach).
> 
> If you still need me to test using RC1, I should be able to get to it later
> today.

Thanks for the info.  Don't bother testing RC1; the reproduction issues were due to the fact that I did not know Crystal was required to make this bug appear.

Tim

(In reply to Timothy Pearson from comment #21)
> (In reply to Kristopher from comment #20)
> > I am using Crystal with transparency enabled in Crystal's background
> > settings. I have also enabled translucency under Control Center -> Window
> > Behavior -> Translucency. If you need more specific details, let me know (or
> > just point me to the config files to attach).
> > 
> > If you still need me to test using RC1, I should be able to get to it later
> > today.
> 
> Thanks for the info.  Don't bother testing RC1; the reproduction issues were
> due to the fact that I did not know Crystal was required to make this bug
> appear.
> 
> Tim

Confirming that the bug is related to transparency; disabling transparency resolves the issue.

Now to figure out why transparency would be affecting window states...my best guess is the DCOP call added in this commit:
https://git.trinitydesktop.org/cgit/twin-style-crystal/commit/?id=8095005fe253b65fa87f48008db7560c7bc0eaac

This came down to not being able to use the TWinModule class in a window decorator; fixed in GIT hash 80aec91.

Please test and confirm; this will also be part of R14 RC2.

(In reply to Timothy Pearson from comment #23)
> This came down to not being able to use the TWinModule class in a window
> decorator; fixed in GIT hash 80aec91.
> 
> Please test and confirm; this will also be part of R14 RC2.

After doing a dist-upgrade (which included twin) and rebooting earlier this morning, the only thing I can confirm is that we seem to have gone "backwards" in getting this fixed.

The taskbar no longer updates upon switching tasks and often directly resembles the screenshot I attached. Keyboard shorcuts don't work anymore except for the task with a bold taskbar icon. Translucency effects do not seem applied to inactive windows like they should be, though I doubt this is an issue with the translucency code as the settings seem to apply correctly when I move around windows.

While twin was already rebuilt the crystal style (which is packaged separately) has not yet been rebuilt.  As the fundamental error was within the crystal style, not twin, your results are not surpising.

Please test after you receive a twin-style-crystal update.

Thanks!

(In reply to Timothy Pearson from comment #25)
> While twin was already rebuilt the crystal style (which is packaged
> separately) has not yet been rebuilt.  As the fundamental error was within
> the crystal style, not twin, your results are not surpising.
> 
> Please test after you receive a twin-style-crystal update.
> 
> Thanks!

The twin-style-crystal update should now be available.  Please test to confirm that the bug is now resolved.

Thanks!

The Crystal update finally came through, and as far as I can see everything described in this report and it's comments seems to be fixed. I am thus marking it as RESOLVED FIXED .

(In reply to Kristopher from comment #27)
> The Crystal update finally came through, and as far as I can see everything
> described in this report and it's comments seems to be fixed. I am thus
> marking it as RESOLVED FIXED .

Great!  Thanks for testing, and sorry for the delay in figuring out what was actually causing this.

(In reply to Timothy Pearson from comment #28)
> (In reply to Kristopher from comment #27)
> > The Crystal update finally came through, and as far as I can see everything
> > described in this report and it's comments seems to be fixed. I am thus
> > marking it as RESOLVED FIXED .
> 
> Great!  Thanks for testing, and sorry for the delay in figuring out what was
> actually causing this.

For me, it also explained why I did not notice this bug on my test machine.