On Saturday 29 of December 2007, Maksim Orlovich wrote: > > Hi, > > > > I'm glad that a KDE dev is finally taking an active interest in XEmbed > > for KDE/Konqueror: > > > > http://www.kdedevelopers.org/node/3162 > > > > Lubos? Are you out there? If you contact me, maybe we can figure out how > > the KDE/XEmbed <-> Flash interaction is going bad. And you'll be a hero > > to KDE users everywhere. > > Thank you for contacting us. > > I am CCing him on this message, I wasn't subscribed to kfm-devel anymore, but I've subscribed again for the time being. > he is more of an X guy though (and hence > more able of dealing with event loops and such), I would be the other > contact, for the KDE4 version. > > In my version, I am seeing 2 problems: > 1. When NPN_UserAgent returns our real UA string, on the cleanup call, > flash calls XtRemoveTimeout with a bogus value; if I workaround that it > crashes with some memory corruption. E.g. http://bugs.kde.org/show_bug.cgi?id=154462 has a backtrace of this, not that it would be very useful, with no debug info. Valgrind shows loads of problems inside of Flash, but nothing that seems to be relevant to this. Using UA string of Firefox avoids this, so this is rather clearly a Flash bug. > 2. Stuff just randomly stalls. I really have no clue on how to debug that, > unfortunately. (Will try DiamondX, just need some help to register its > weird mimetype). I cannot confirm this one, I don't remember noticing this. BTW, for DiamondX, rename the binary to 'libdiamondx.so' (i.e. dump the numbers) and put it wherever your other plugins are, that makes it work here. 3. Keyboard focus doesn't work - e.g. at http://flashgames247.com/play/859.html, holding the Up arrow doesn't do anything, but it works with Firefox or r48. DiamondX reports focus in/out events on clicks, so presumably a Flash bug too (especially if it still has the ugly keyboard handling that led to the security problem a year ago). 4. Clicking the flash at http://www.nintendo.de does not work. Again, works fine with r48 or Firefox. Clicking otherwise seems to work though. 5. Repeated calls to NPP_SetWindow() are not handled properly by r115. Konqueror first creates the plugin window and then (during KHTML layouting) it resizes it, which, e.g. according to http://devedge-temp.mozilla.org/library/manuals/2002/plugin/1.0/npp_api12.html, results in two NPP_SetWindow() calls. While r48 seems to handle them fine, r115 creates a new flash object on each call instead of just adjusting the already existing one. 6. People report random crashes, e.g. https://bugzilla.novell.com/attachment.cgi?id=188041, I cannot reproduce though. I'd like to also point out that the DiamondX plugin seems to work without problems with Konqueror and I don't see any important difference in its debug output when compared to when running in Firefox. Current SVN should have my latest code in the 3.5 branch, both latest kdelibs and kdebase are needed. Some attempts at workarounds for some of the issues (1,5) are included. Distributions are already shipping updates, but those may not be my latest patches. -- Lubos Lunak KDE developer -------------------------------------------------------------- SUSE LINUX, s.r.o. e-mail: l.lunak@suse.cz , l.lunak@kde.org Lihovarska 1060/12 tel: +420 284 028 972 190 00 Prague 9 fax: +420 284 028 951 Czech Republic http//www.suse.cz