RFR: 8332106: VerifyError when using switch pattern in this(...) or super(…)
Jan Lahoda
jlahoda at openjdk.org
Mon May 13 14:43:01 UTC 2024
Considering code like:
public class Test {
public Test(int i) {
}
public Test(Object o) {
this(o instanceof R(String s) ? s.length() : -1);
}
public static void main(String... args) {}
record R(String s) {}
}
Running this crashes:
$ java .../Test.java
Exception in thread "main" java.lang.VerifyError: Stack map does not match the one at exception handler 36
Exception Details:
Location:
Test.<init>(Ljava/lang/Object;)V @36: astore_2
Reason:
Type uninitializedThis (current frame, locals[0]) is not assignable to 'Test' (stack map, locals[0])
Current Frame:
bci: @14
flags: { flagThisUninit }
locals: { uninitializedThis, 'java/lang/Object', top, 'Test$R' }
stack: { 'java/lang/Throwable' }
Stackmap Frame:
bci: @36
flags: { }
locals: { 'Test', 'java/lang/Object' }
stack: { 'java/lang/Throwable' }
Bytecode:
0000000: 2a2b c100 0799 0018 2bc0 0007 4e2d b600
0000010: 093a 0419 044d 2cb6 000d a700 0402 b700
0000020: 13a7 0011 4dbb 001a 592c b600 1c2c b700
0000030: 1fbf b1
Exception Handler Table:
bci [14, 17] => handler: 36
Stackmap Table:
same_locals_1_stack_item_frame(@29,UninitializedThis)
full_frame(@30,{UninitializedThis,Object[#2]},{UninitializedThis,Integer})
full_frame(@36,{Object[#20],Object[#2]},{Object[#24]})
same_frame(@50)
at java.base/java.lang.Class.forName0(Native Method)
at java.base/java.lang.Class.forName(Class.java:534)
at java.base/java.lang.Class.forName(Class.java:513)
at jdk.compiler/com.sun.tools.javac.launcher.Main.execute(Main.java:432)
at jdk.compiler/com.sun.tools.javac.launcher.Main.run(Main.java:208)
at jdk.compiler/com.sun.tools.javac.launcher.Main.main(Main.java:135)
The reason is that there is a synthetic catch generated wrapping the record accessors inside the pattern matching. The range where the exception this catch is catching is before the real `this(...)` `invokespecial`, and hence the `this` is still an "uninitialized this". But the code for the catch is generated at the end of the constructor, when `this` is already initialized, and javac generates the stack maps as if `this` was initialized.
In general, the pattern matching code can be both before and after the `this` has been initialized. But, I don't think a stack map frame can be generated for the handler originating in places with both "uninitialized this " and (initialized) this.
The proposal here is to generate one set of the catch handlers for the state with "uninitialized this", and another for (initialized) this, if needed.
-------------
Commit messages:
- 8332106: VerifyError when using switch pattern in this(...) or super(...)
Changes: https://git.openjdk.org/jdk/pull/19217/files
Webrev: https://webrevs.openjdk.org/?repo=jdk&pr=19217&range=00
Issue: https://bugs.openjdk.org/browse/JDK-8332106
Stats: 320 lines in 3 files changed: 305 ins; 4 del; 11 mod
Patch: https://git.openjdk.org/jdk/pull/19217.diff
Fetch: git fetch https://git.openjdk.org/jdk.git pull/19217/head:pull/19217
PR: https://git.openjdk.org/jdk/pull/19217
More information about the compiler-dev
mailing list