During the investigation of 5089429 we've run into another problem caused by
mutlithread access to a Graphics object.
Run test/java/awt/Graphics2D/MTGraphicsAccessTest with -full parameter
(a multiple-cpu system is preferred), on a jdk with fixed 5089429 and the
following crash may occur:
current thread: t@33
  [1] _lwp_kill(0x0, 0x6, 0x0, 0xff33c000, 0x0, 0x0), at 0xff31e42c
  [2] raise(0x6, 0x0, 0xd3efed48, 0x7efefeff, 0x81010100, 0xff00), at 0xff2ccd70
  [3] abort(0xd3efedf0, 0xfec9c756, 0xfec9c753, 0x21, 0x81010100, 0xff00), at 0xff2b5c60
  [4] os::abort(0x1, 0xfed2c25c, 0x7d0, 0xfee53078, 0x0, 0x0), at 0xfe77af28
  [5] VMError::report_and_die(0xd3efefc8, 0x20caa8, 0xfebe862e, 0xfebe85f8, 0x4f, 0x0), at 0xfea702a8
  [6] report_fatal(0xfebe85f8, 0x4f, 0xfebe862e, 0x148b, 0x9c2db90, 0xd3eff450), at 0xfe1708f0
  [7] jfieldIDWorkaround::verify_instance_jfieldID(0xd619ce50, 0xf9328833, 0x4, 0xfa8294a0, 0x25c, 0xd3eff4e0), at 0xfe397c38
  [8] jfieldIDWorkaround::from_instance_jfieldID(0xd619ce50, 0xf9328833, 0x4, 0xfa829060, 0xefa5ec78, 0xd3eff588), at 0xfe3d398c
  [9] jni_GetIntField(0x20cb74, 0x2129e4, 0xf9328833, 0xfa8297c0, 0xefa5ec78, 0xd3eff620), at 0xfe3b6fe0
=>[10] GrPrim_CompGetXorInfo(env = 0x20cb74, pCompInfo = 0xd3eff2ac, comp = 0x2129e4), line 438 in "GraphicsPrimitiveMgr.c"
  [11] GrPrim_Sg2dGetCompInfo(env = 0x20cb74, sg2d = 0xd3eff4c0, pPrim = 0xd5d2f710, pCompInfo = 0xd3eff2ac), line 390 in "GraphicsPrimitiveMgr.c"
  [12] Java_sun_java2d_loops_FillSpans_FillSpans(env = 0x20cb74, self = 0xd3eff4c4, sg2d = 0xd3eff4c0, sData = 0xd3eff4bc, pixel = 2139094784, pIterator = -707586392LL, si = 0xd3eff4ac), line 44 in "FillSpans.c"

  It looks like we started rendering in xor mode, but the sg2d.composite
  got changed by another thread before we pulled it from the native code,
  so we then attempted to get the xor pixel color from a different
  composite..

###@###.### 2004-12-08 00:20:48 GMT