Mercurial > hg > Members > kono > jpf-core
comparison src/main/gov/nasa/jpf/jvm/bytecode/NATIVERETURN.java @ 27:8aded593a50f
fixed the missing class init status update for native clinits. Since we moved
that to the clinit RETURN in r24, it has to be duplicated in NATIVERETURN for native
clinits. Factoring out to DIRECTCALLRETURN is not an option since there is no
good way to determine if the direct call included a clinit (we would have to
look at the code, which is suboptimal performance- and change resilience-wise).
Thanks to Franz Weitl for the report
| author | Peter Mehlitz <pcmehlitz@gmail.com> |
|---|---|
| date | Mon, 11 May 2015 12:17:18 -0700 |
| parents | 61d41facf527 |
| children |
comparison
equal
deleted
inserted
replaced
| 26:4eb191cbb68c | 27:8aded593a50f |
|---|---|
| 57 | 57 |
| 58 if (retAttr != null) { | 58 if (retAttr != null) { |
| 59 setReturnAttr(ti, retAttr); | 59 setReturnAttr(ti, retAttr); |
| 60 } | 60 } |
| 61 | 61 |
| 62 if (mi.isClinit()) { | |
| 63 // this is in the clinit RETURN insn for non-MJIs so we have to duplicate here | |
| 64 // Duplication could be avoided in DIRECTCALLRETURN, but there is no reliable | |
| 65 // way to check if the direct call did return from a clinit since the corresponding | |
| 66 // synthetic method could do anything | |
| 67 mi.getClassInfo().setInitialized(); | |
| 68 } | |
| 69 | |
| 62 return frame.getPC().getNext(); | 70 return frame.getPC().getNext(); |
| 63 } | 71 } |
| 64 | 72 |
| 65 @Override | 73 @Override |
| 66 public void cleanupTransients(){ | 74 public void cleanupTransients(){ |