[oi-dev] X11 update

[ken mays]

Martin,
Well, you detailed out what I was aiming at - it is not the driver. Yes, I know about the Solaris graphics backend components as it was discussed years ago - straight from the Sun USA/China kernel teams and X11 team. I'vewalked the green mile with each graphics driver whether from Sun, IHV, or open source. The more we talk about this the older I get.

Your point is clear - don't use Intel SNA for Solaris/illumos based systems at this time. I didn't suggest it as it'll make your hair turn silvery grey! 
They used to say "love is all you need" ... guess not.
~K
 



    On Friday, September 16, 2016 5:52 PM, Мартин Бохниг <opensxce at mail.ru> wrote:
 

 
> Kernel DRM needs more love.

Unfortunately I spent all of this year for Hipster related projetcs.
It doesn't need "more love" but some $$$ to pay the running bills, but I understand that community work will never ever be sustainable to us here who actually do something (and I'm only one of many), hence forget my comment.
Vbox 5.1.6 is almost ready to ship, I made it work with Qt4.8 aagain after having messsed 1 week with Qt5.8alpha (and other 5.x branches).
Qt 5.x needs more love, and its newer versions are not yet compatible with Vbox even *if* and *after* you got it all built for 64bit (and Vbox is 64bit-only since 5.0).


As of Intel DRM versus SNA:

Sun and then Oracle never enabled SNA for their Xorg ddx, they strictly defaulted and still keep defaulting to UXA.
And there is a good reason for this: SNA performs about 10 times *slower* on Solaris kernels than with accelleration disabled in the Intel ddx!
You can test that on Intel Xorg ddx 2.2x with SNA enabled. Unlike 2.99.x which is a _pre_release 3.0 _non_version (as shipped in Hipster), the 2.2x Intel ddx branch does _not_ segfault with the S12 DRM/KMS bakport. But you never in your life witnessed such a sloooooowwwww-mooootion anti-performance if you explicitly make it use SNA (by compiling in SNA support _plus_ [on Solars] by explicitly specifying SNA accell via xorg.conf).

After you experienced that (I mentioned it last winter) you will stop whining abbout lack of SNA.
Reminder: Sun/Oracle also intentionally do NOT use it.

So rather than spreading such vague uncertainty like "needs more love", be more specific: The only thing which i915/drm-kms works on Hipster is support for <= generation 5 Intel GPU's.
And to get *that*, not the driver needs any single line of change, but Illumos' outdated agpgart stack needs to get enhanced.
And that would already have happened had I not wasted 1 week on Qt5 which is really a pain on my Hipster installation based on illumos-f83b46b .

Because even after converting Aurelien's specs to match qt5/qt-everywhere-opensource-src-5.8.0-alpha/qtbase/mkspecs/solaris-g++-64 on my machine it still didn't work at all with Qt5.6 or higher, due to changes in qmake.
I always ended up in this exact scenario: http://stackoverflow.com/questions/38549755/raspberry-pi-2-and-qt-5-7-make-stuck-at-qtbase-bin-qmake-conf-qtbase-q 

And that also includedes the workaround.
However, with the resulting painfully built Qt5.8 and based on that Vbox linked against it all one would get was 20 lines of *this*:  "gtk_menu_attach_to_widget(): menu already attached to GtkMenuItem"

KMS: Needs more love: What arou you waiting for?
Specifics?
diffs?


%m



 Пятница, 16 сентября 2016, 23:25 UTC от ken mays via oi-dev <oi-dev at openindiana.org>:3. On Intel, UXA works well while SNA still crashes.
- SNA is improved in current Intel drv git. Kernel DRM needs more love. 

~K


Unfortunately I spent all of this year for Hipster related projetcs.


   
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://openindiana.org/pipermail/oi-dev/attachments/20160917/5f25a91d/attachment-0005.html>