| [#242] | project: compiler | priority: low | category: bug | |
|---|---|---|---|---|
| submitter | assigned to | status | date submitted | |
| Burak | Michel | fixed | 2003-12-01 12:38:55.0 | |
| subject | long patterns crash backend | |||
| code |
object bugFJ_2 {
def extract( cs:char* ):Unit = cs match {
case Seq(_*,'\\',src @ _ *,'\\','e','n','d','{','l','s','t','l',rest @ _ * )
=>
}
}
|
|||
| what happened | /home/buraq/src/scala/exciss> scalac exciss.scala Exception in thread "main" java.lang.ArrayIndexOutOfBoundsException: -21657 at ch.epfl.lamp.fjbg.JCode.computeMaxStackSize(JCode.java:814) ... at ch.epfl.lamp.fjbg.JCode.getMaxStackSize(JCode.java:86) at ch.epfl.lamp.fjbg.JCodeAttribute.writeContentsTo(JCodeAttribute.java:77) at ch.epfl.lamp.fjbg.JAttribute.writeTo(JAttribute.java:61) at ch.epfl.lamp.fjbg.JAttribute.writeTo(JAttribute.java:25) at ch.epfl.lamp.fjbg.JFieldOrMethod.writeTo(JFieldOrMethod.java:133) at ch.epfl.lamp.fjbg.JClass.writeTo(JClass.java:320) at ch.epfl.lamp.fjbg.JClass.writeTo(JClass.java:272) at ch.epfl.lamp.fjbg.JClass.writeTo(JClass.java:256) at scalac.backend.jvm.GenJVM.leaveClass(GenJVM.java:1563) ... |
|||
| what expected | no error | |||
| [back to overview] | ||||
| Michel edited on 2003-12-02 09:27:29.0 |
|
This particular instance of the bug was fixed by adding support for wide jumps in the back-end. All methods are first compiled with short jumps, and if it happens that at least one jump needs a long offset, the compilation of the whole method is restarted with wide jumps all over the place. The general problem, as discussed with Buraq, is that the pattern matcher generates huge code, which one day or another is bound to break the (small) limits of the JVM. We should find a solution for that, but I don't know yet whether we want to do it in the back-end or earlier. |
| Burak edited on 2003-12-02 16:51:09.0 |
| same as for bug 243, I changed the inefficient translation in the TransMatch phase. |