[aarch64-port-dev ] TestStackBangRbp hangs

[Cao Hoang Thu]

Hi Ed,

With C2 on RTSM --> hang indefinitely all test-cases

With C1 on RTSM-->
java -Xcomp TestStackBangRbp -> hang 
java -Xcomp -XX:-DeoptimizeWhenPatching TestStackBangRbp --> test passes

Regards,
Thu Cao

--------------------------------------------
On Wed, 1/8/14, Edward Nevill <edward.nevill at linaro.org> wrote:

 Subject: [aarch64-port-dev ] TestStackBangRbp hangs
 To: "aarch64-port-dev at openjdk.java.net" <aarch64-port-dev at openjdk.java.net>
 Date: Wednesday, January 8, 2014, 6:18 PM
 
 Hi,
 
 The following test program causes openjdk to hang
 indefinitely requiring a kill -9 to stop the process.
 
 To replicate
 
 - compile the test program
 
 javac TestStackBangRbp.java
 
 - run it as follows
 
 java -Xcomp TestStackBangRbp
 
 Note that the problem occurs for both client and server and
 also occurs on both builtin sim and on the RTSM model.
 
 If you add
 
 java -Xcomp -XX:+PrintCompilation -XX:+TracePatching
 TestStackBangRbp
 
 the last few lines of the output look like
 
    8291  368    b   
     TestStackBangRbp::<clinit> (86 bytes)
 Deoptimizing because patch is needed
    8302  368       
      TestStackBangRbp::<clinit>
 (86 bytes)   made not entrant
    8303  369    b   
     TestStackBangRbp::main (38 bytes)
    8341  370    b   
     TestStackBangRbp::m1 (246 bytes)
    8375  371   !b 
       TestStackBangRbp::m2 (35 bytes)
 Deoptimizing because patch is needed
    8463  371   ! 
        TestStackBangRbp::m2 (35
 bytes)   made not entrant
 
 If I then add -XX:-DeoptimizeWhenPatching, IE
 
 java -Xcomp -XX:-DeoptimizeWhenPatching TestStackBangRbp
 
 The test passes.
 
 So it looks like a problem with stack overflow when
 deoptimizing (or having just deoptimized). 
 
 Looks like fun:-).
 
 Ed.