QT overrides several of the xcursor cursors from XFree86 4.2.99 and greater. This can cause things to be really slow unles you know how to turn them off. But why turn them off when you can compile QT to let them in! Reproducible: Always Steps to Reproduce: 1. Install xfree86 4.2.99 2. run kde Actual Results: cursor disappear for several seconds when you move over icons or on window borders unless you've enabled XCURSOR_CORE=true Expected Results: used the nice xcursors :)
Created attachment 8005 [details, diff] patch to add xcursor option to configure This goes with the ebuild I wrote to optionally add xcursor support
not sure why the patch upload didn't work. If the developer emails me I'll send both patches via email.
Created attachment 8057 [details, diff] cursor.diff from faye
Created attachment 8058 [details, diff] qt-3.0.1-r3 diff also from faye
OK, the switch between normal and new cursors shouldn't be slow. Workaround add Option "SWcursor" to override accel feature being used. Replacement DRI module in kernel fixes the problem. All this patch does is let you use more pretty cursors, which is a good thing.
hmm, my mouse cursor does work without this patch. for which mouse cursors is this needed? if this patch makes sense, should we add myconf="${myconf} -xcursor" by default to qt ebuild? or will this break xfree-4.2.x?
basically QT overrides several X cursors which means that these black/white ones get used in place of the pretty XCURSOR extension ones. All the patch does is enable the USE_APPROXIMATE_CURSORS (IIRC) define which stops it from overriding, for example, the resize cursors, if you go to resize a window without this patch you get the old black/white versions. It's non destructive to enable it on any system, but I wouldn't do it by default as the QT cursors are quite nice unless you've got better ones to use :)
"basically QT overrides several X cursors which means that these black/white ones get used in place of the pretty XCURSOR extension ones." Qt offers more cursors then xfree. AFAIK what you described happens because Qt uses sometimes a cursor which does not exist in a given cursor theme, but links can help. Links which the Gentoo cursor themes are usually missing are: 00008160000006810000408080010102 -> v_double_arrow 028006030e0e7ebffc7f7070c0600140 -> h_double_arrow 03b6e0fcb3499374a867c041f52298f0 -> crossed_circle 08e8e1c95fe2fc01f976f1e063a24ccd -> left_ptr_watch 14fef782d02440884392942c11205230 -> sb_h_double_arrow 2870a09082c103050810ffdffffe0204 -> sb_v_double_arrow 3ecb610c1bf2410f44200f48c40d3599 -> left_ptr_watch 4498f0e0c1937ffe01fd06f973665830 -> left_ptr 9d800788f1b08800ae810202380a0822 -> hand1 c7088f0f3e6c8088236ef8e1e3e70000 -> bd_double_arrow d9ce0ab605698f320427677b458ad60b -> question_arrow e29285e634086352946a0e7090d73106 -> hand2 fcf1c3c7cd4491d801f1e1c78f100000 -> fd_double_arrow (I have these names from a buried post on kde-look.org) In this context I could imagine that XCURSORS_CORE=true forces Qt to use X11 core pointers only. But I haven't checked this and it's only speculation from my part.
Hmm, the interesting bugs everyone forgot about (and that I never had anything to do with)... Updates please; is this still relevant with current qt and x.org versions? Besides, according to comment #8, qt doesn't need to be patched; rather this bug should be redirected towards the cursor theme people. (Another thing I'm not involved with, so can't comment - I never used XCURSOR to begin with, I'm just cleaning up old bugs.)
This doesn't apply to newer versions of qt, I guess. Closing.
