Apache OpenOffice (AOO) Bugzilla – Issue 107249
More X11 fullscreen mode changes
Last modified: 2017-05-20 10:23:45 UTC
Sooo - pretty much as everyone expected, my dock/toolbar swap in issue 103145 causes a regression - metacity in some very specific dual monitor xinerama setups (seems to be related to monitor offsets, sizes or somesuch) puts all fullscreen windows (i.e. slideshow & presenter console) on the same monitor. Chatted a bit with Lubos, who suggested the kwin _NET_WM_STATE_FULLSCREEN issue we had all those workarounds in place is long since gone - so now stripped down the code to use explicit positioning of the window + _NET_WM_STATE_FULLSCREEN instead, tested on metacity, compiz, kwin & xfce and everything seems (again) to work flawlessly. For really bleeding-edge distros, additionally using http://standards.freedesktop.org/wm-spec/wm-spec-latest.html#id2552578 could be useful (not in my fix). Patch is work in progress, there's currently e.g. still a difference between gtk & gen plug, in that for gtk the fullscreen windows stay background-able (which I find rather helpful, e.g. when doing some demo stuff).
Created attachment 66367 [details] A patch
n.b. the difference is of course due to the left-over GetDisplay()->getWMAdaptor()->enableAlwaysOnTop() call - did my tests on opensuse 11.0 (metacity-2.22.0, compiz-0.7.4, kde4-kwin-4.0.4, xfce4-4.4.2), and 11.2.
*** Issue 110482 has been marked as a duplicate of this issue. ***
@pl: so with issue 110482, should we move forward with this for 3.3? Conceptually, I'd like to back off from too many wm-hackarounds, like just using STATE_FULLSCREEN here. If anything, then add specific hacks conditional to specific wm versions. What do you think?
I'd be more than happy to (I really hate to work around window managers getting in the way of things, I have enough bugs of my own to care about), only that I suspect "long since gone" is a timeline of at max a year, right ? Our baseline is a tad older than that however. I suspect this would be breaking older distributions in favor of newer ones.
@pl: hard to gauge, really. I tried it on an opensuse 11.0, which is old under pretty much any standard, and it works. OTOH, that's why I thought about having conditional hacks for specific WM versions - have to poke Lubos, but I'm pretty sure his notion of old and ours coincide.
Yes, it's really a matter of perspective :-) For what it's worth, personally I use OpenSuSE 11.2. But our baseline at http://wiki.services.openoffice.org/wiki/Compiler_versions_used_by_port_maintainers_and_release_engineers indicates ancient things like SuSE 9.2. Personally I'd have to dig some time to find installation CDs for something so old, but there it is.
Since this patch would probably kill two birds with one stone (see issue 110881) I see more urgency for this. Actually I'd even not mind a 3.2.1 target because the presenter console is kind of important to me as to not getting hit with a stick by af. Only I'd like to have the current code on older versions, so maybe we can query metacity, kwin and compiz for their version. I'll see that now.
nominated for 3.2.1, I intend to commit the (enhanced version) of this patch to CWS ooo321gsl05
done so, changeset at http://hg.services.openoffice.org/hg/cws/ooo321gsl05/rev/a03d0eb91d8d
The only version check currently however is for metacity, compiz was not obliging in that regard. Well that problem should solve itself over time when our baseline goes up.
please verify in CWS ooo321gsl05
Verified in CWS.
*** Issue 96919 has been marked as a duplicate of this issue. ***
*** Issue 112276 has been marked as a duplicate of this issue. ***