rdesktop workspace disappears ocassionally

[GoogleCodeExporter]

What steps will reproduce the problem?
1. Make a fullscreen remote desktop session and give it a unique tag
("remote").
2. Exit fullscreen (Ctrl-Alt-Enter) and change workspaces.  The "remote"
workspace should have disappeared from the "taskbar".  If not, repeat a few
times and it surely will.

What is the expected output? What do you see instead?
It should not disappear from the taskbar.  In fact, it still exists, the
name just doesn't show up on the taskbar.  If I ALT-P and type "remote", it
still pulls up.

What version of the product are you using (wmii -v)? On what operating system?

wmii-3.6 on Arch Linux

Original issue reported on code.google.com by sch...@gmail.com on 22 Jun 2009 at 4:04

[GoogleCodeExporter]

[deleted comment]

[GoogleCodeExporter]

Silly me, I meant 'ALT-T and type "remote"', not ALT-P.  When typing "remote",
"remote" is not in the list and there is not autocomplete.  It is as if it 
doesn't
exit, but in the end it does!

Original comment by sch...@gmail.com on 22 Jun 2009 at 4:06

[GoogleCodeExporter]

Hmmm... it turns out that rdesktop shows up on ANY workspace that does not have
anything else.  For example, if I switch to workspace "remote" I see rdesktop 
and
also if I switch to workspace "asdf".

Original comment by sch...@gmail.com on 22 Jun 2009 at 4:54

[GoogleCodeExporter]

Sorry, I thought I'd commented on this before. The rdesktop program is very 
hacky. I
haven't been able to reproduce this, but as I recall, rdesktop uses some very 
ugly
hacks to enable fullscreen mode (as does vncviewer).

Original comment by maglion...@gmail.com on 2 Oct 2009 at 4:43

[GoogleCodeExporter]

I agree.  I've seen this problem as well.  As a workaround, I use rdesktop as a
floating client whose dimensions are that of my screen (1920x1200) minus space 
for
wmii's status bar.

rdesktop remote_host -g 1915x1155 -a 16 -z

Original comment by sun...@gmail.com on 2 Oct 2009 at 10:04

• Added labels: Component-Logic, Type-Defect

[GoogleCodeExporter]

You can also use M-f (or wmiir write /client/sel/ctl Fullscreen on), which 
rdesktop
doesn't know or care about. The only trick is triggering the mouse/keyboard grab
(which my wmiirc has a special mode for).

Original comment by maglion...@gmail.com on 2 Oct 2009 at 10:10

[GoogleCodeExporter]

How do you set this special mode for M-f in your wmiirc?

Original comment by sch...@gmail.com on 2 Oct 2009 at 10:32

[GoogleCodeExporter]

It doesn't really have anything to do with M-f; M-f is just a non-hacky way to 
set a
window fullscreen. I do have a special view, though, in which all wmii key 
bindings
but one are enabled. I use it for testing wmii. Mod-x toggles to and from it.

tags.ignore.add('x')
keys.bind('main',   { '%(mod)s-x': lambda k: tags.select('x') })
keys.bind('xembed', { '%(mod)s-x': lambda k: tags.select(tags.PREV) })
bind_events({
    Match('FocusTag', 'x'):   lambda *a: setattr(keys, 'mode', 'xembed'),
    Match('UnfocusTag', 'x'): lambda *a: setattr(keys, 'mode', 'main'),
})

I'm closing this issue as invalid, since it appears to be a rdesktop bug rather 
than
a wmii bug.

Original comment by maglion...@gmail.com on 6 Oct 2009 at 8:24

• Changed state: Invalid

[GoogleCodeExporter]

I ran rdesktop from a terminal today and noticed that it emits the following 
warnings
whenever the problem described in this bug report occurs:

WARNING: Broken Window Manager: doesn't handle restack (restack request was 
ignored)

Original comment by sun...@gmail.com on 6 Oct 2009 at 10:14

[GoogleCodeExporter]

It should rather read "Broken X11 client". ICCCM doesn't require that window
managers handle restack requests.

    Window managers are in any case free to position windows in the stack as
    they see fit, and so clients should not rely on receiving the stacking
    order they have requested.

The warning is irrelevant to the issue, though. The problem is that rdesktop 
chooses
to use override-redirect windows --- windows that wmii is supposed to ignore 
--- to
implement its fullscreen mode. This is absurd, given that there are already
mechanisms to handle fullscreen windows, and rdesktop apparently has some EWMH
support anyway. Moreover, ICCCM specifies that override-redirect windows should 
be
short lived; it's mainly meant for popups and menus. Regardless, given 
rdesktop's
tack, it's its own job to handle its own stacking order (wmii doesn't even see 
its
restacking requests at that point). 

Original comment by maglion...@gmail.com on 6 Oct 2009 at 10:56

[GoogleCodeExporter]

Try the attached rdesktop patch. It should solve the problem,
but Knuth's adage applies,

    Beware of bugs in the above code; I have only proved it
    correct, not tried it.

It builds, but as I don't have access to an rdesktop server, I
can't promise that it actually creates a working fullscreen
window.

Original comment by maglion...@gmail.com on 6 Oct 2009 at 11:17

Attachments:

• ewmh-fullscreen.patch

[GoogleCodeExporter]

Well, this is embarrasing.

I compiled rdesktop 1.6.0 from source and ran it -- just to make sure I could
reproduce the problem before applying your patch -- and, the problem did NOT 
occur! o_O

I applied your patch and compiled rdesktop and it crashed.  ;-)
(Sorry for wasting your time trying to create that patch.)

In conclusion, it seems that Debian/Ubuntu's additional patches on top of the 
stock
rdesktop 1.6.0 sources that are creating the problem.  >_<

Original comment by sun...@gmail.com on 7 Oct 2009 at 12:53