Hi Sergey, > Did you check what sequence of calls causes a situation when we have > BufImgSurfaceData as a surface and XRRenderer as a Pipe?
No. The issue is really difficult to reproduce and it disappears with any logging enabled, so it might be a race condition. I was able to reproduce it artificially by resizing VMware VM number of times with high frequency. It also was reported by users on KDE desktops. Best Regards, Alexey > On 23 May 2016, at 20:11, Sergey Bylokhov <sergey.bylok...@oracle.com> wrote: > > Hi, Alexey. > Did you check what sequence of calls causes a situation when we have > BufImgSurfaceData as a surface and XRRenderer as a Pipe? I thought we always > change them at once(actually the pipe usually updated from the > surface.validatePipe(SG2D))? > > On 23.05.16 17:44, Alexey Ushakov wrote: >> Hello Phil, >> >> Here is a small fix of quite annoying exception that we sometimes have >> in IDEA product. The bug is assigned to you so, please have a look. >> >> Bug: https://bugs.openjdk.java.net/browse/JDK-7179454 >> Webrev: http://cr.openjdk.java.net/~avu/JDK-7179454/webrev.00 >> >> Sometimes rendering is performed into surface of wrong type, so we need >> to recreate it by throwing InvalidPipeException. The exception is caught >> at upper level and the surface of appropriate type is created. Similar >> approach is used in OpenGL and D3D pipeline. >> >> Best Regards, >> Alexey > > > -- > Best regards, Sergey.
