[kaffe] ThreadLocalTest failure in night tests

[Patrick Tullmann]

FYI, an intermittent bug in Kaffe cropped up in last night's nightly
test suite.  (I say its intermittent because it didn't occur last
night, and I don't think Kaffe was changed yesterday.)

> - Real tests for configopt "--with-engine=jit3 --disable-debug --with-staticlib --with-staticvm"
> 
...
> FAIL: ThreadLocalTest.java
>    accessing free page 0xdad65ca8 (above 0x8176000)
>    x I0	 TL:  TL initial
>    x I0	ITL: ITL initial
>    x I1	 TL:  TL initial changed by x at k == 0
>    x I1	ITL: null
>    y I0	 TL:  TL initial
>    y I0	ITL: ITL initial
>    y I1	 TL:  TL initial changed by y at k == 0
>    y I1	ITL: null
>    y I2	 TL:  TL initial changed by y at k == 0 changed by y at k == 1
>    y I2	ITL: set by y at k == 1
>    y I3	 TL:  TL initial changed by y at k == 0 changed by y at k == 1 changed by y at k == 2
>    y I3	ITL: null

The output for this test is sorted, so its not clear when the bad
access occured.  Anyone recognize 0xdad65ca8 as magic digits of any
sort?  (Its not ascii.)

If anyone feels like playing around with this, there is a 'repeatrun'
macro in developers/gdbinit that might help.  Note the static linkage
in the configuration.

-Pat

----- ----- ---- ---  ---  --   -    -      -         -               -
Pat Tullmann                                       tullmann@cs.utah.edu
     To understand recursion one must first understand recursion.