On Sun, 18 Oct 2020 03:36:27 GMT, Sergey Bylokhov <s...@openjdk.org> wrote:
>> Please review a test fix for an issue seen failing in mach5 testing although >> it is not reproducible locally. >> The issue seems to stem from the fact that the test was been run as headless >> test although it requires to create UI >> class like ComponentUI which is needed for painting, layout calculations etc >> and therefore it fails citing >> _no ComponentUI class for: >> javax.swing.JButton[,0,0,0x0,invalid,alignmentX=0.0,alignmentY=0.0,border=,flags=0,maximumSize=,minimumSize=,preferredSize=,defaultIcon=,disabledIcon=,disabledSelectedIcon=,margin=null,paintBorder=true,paintFocus=true,pressedIcon=,rolloverEnabled=false,rolloverIcon=,rolloverSelectedIcon=,selectedIcon=,text=Probably,defaultCapable=true]_ >> Modified the test to mark it as headful. >> Mach5 job was run for several iteration on all 3 platforms. Link in JBS > > It does not look like the test uses any API which required the headful > system, none of them throw a headless exception. if you run it locally with -Djava.awt.headless=true what happens ? And in the bug report I see Caused by: java.lang.NullPointerException at java.desktop/javax.swing.plaf.basic.LazyActionMap.installLazyActionMap(LazyActionMap.java:64) So installUI failed which is why there's no componentUI but why did we get this NPE that started it ? ------------- PR: https://git.openjdk.java.net/jdk/pull/719
