From mboxrd@z Thu Jan 1 00:00:00 1970 From: Gary Bisson Date: Wed, 8 Mar 2017 12:28:06 +0100 Subject: [Buildroot] QT5.8 buid failure on master In-Reply-To: References: <20170307114826.lend2f2evnob3lya@t450s.lan> <20170307202950.61328dd5@gmx.net> Message-ID: List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: buildroot@busybox.net Hi Jaggan, On Wed, Mar 8, 2017 at 12:00 PM, Jagan Teki wrote: > Hi Garry, > > On Wed, Mar 8, 2017 at 4:15 PM, Gary Bisson > wrote: >> Hi Peter, Fabio, All, >> >> On Tue, Mar 7, 2017 at 8:29 PM, Peter Seiderer wrote: >>> Hello Fabio, >>> >>> On Tue, 7 Mar 2017 13:57:39 -0300, Fabio Estevam wrote: >>> >>>> Hi Jagan, >>>> >>>> On Tue, Mar 7, 2017 at 1:15 PM, Jagan Teki wrote: >>>> >>>> > Able to see the application but 'Black screen' on 'Lord of the Ring' widget >>>> >>>> Thanks for testing. >>>> >>>> I think we should treat this issue on a separate thread as I also see >>>> the same with Qt5.6.2. >>>> >>>> It seems we made progress with Qt5.8 with Peter's patches. >>>> >>>> Peter, >>>> >>>> Do you plan to submit your two patches formally? >>> >>> Yes can do. No problem with the first one, for the second one I need >>> to investigate first on which additional configure option the building >>> of libQt5EglFsKmsSupport.so depends... >> >> Actually I wonder why the first patch is needed in this case. How does >> the eglfs plugin build for Vivante for instance since the failure >> happens in a common file? >> >> Otherwise, I confirm I have "black picture" issues on Qt5.8 with the >> Cinematic demo. >> >> However, I've made a discovery regarding the DRI configuration (to >> avoid hardcoding the device). It is actually documented well on Qt >> website: >> http://doc.qt.io/qt-5/embedded-linux.html#eglfs-with-eglfs-kms-backend >> "The KMS/DRM backend also supports custom configurations via a JSON >> file. Set the environment variable QT_QPA_EGLFS_KMS_CONFIG to the name >> of the file to enable this." >> >> Here is my JSON file for Hannstar display: >> { >> "device": "/dev/dri/card1", >> "hwcursor": false, >> "pbuffers": true, >> "outputs": [ >> { >> "name": "VGA-1", >> "mode": "off" >> }, >> { >> "name": "HDMI-1", >> "mode": "off" >> }, >> { >> "name": "LVDS-1", >> "mode": "1024x768" >> } >> ] >> } >> >> This avoids the patch mentioned before: >> http://lists.busybox.net/pipermail/buildroot/2017-March/185842.html > > Don't know but I never find to explicitly mentioned the kmscube ran > w/o any device argument for 5.6.2 and 5.8 I'm not sure to understand your remark. Note that here I'm only talking about Qt5.8 with Mesa17.0.1. Yes my previous image with Qt5.6.2 and Mesa13.0 from Christian tree didn't require to specify the node. Also, kmscube, which has nothing to do with Qt, now requires you to specify the device (using latest version on master). Previously it was checking the drm name against an "approved" list of devices. Specifying the device manually is therefore more flexible since it doesn't require source code modification to get it working. Finally, I've been able to see the covers of the Cinematic demo using Qt5.8 + Mesa17.0.1 (Etnaviv) by disabling the "Show movies cover lighting" option (bottom left corner). So I guess it's an Etnaviv support issue in Mesa that makes the black rectangles. Regards, Gary