On 2/22/18, 12:51 PM, Xueming Shen wrote:
On 2/22/18, 12:04 PM, Joe Wang wrote:
Hi Sherman,

Thanks for reviewing the change.

Taking a local copy of the count field, but the boundary check would be almost immediately done against the field itself. Are you worrying about the count field may be out of sync with the byte array? I would think that's unlikely to happen. Whether it's StringBuilder or StringBuffer, it's not advisable/practical to use in multiple threads. In a valid usage, the count is always consistent with the byte array.

Hi Joe,

It might not be a "valid usage" but it did happen and when it happens it might just crash the vm without those boundary checks. It's especially true for those intrinsics methods with explicit comments "intrinsic performs no bounds checks". In this case, the StringUTF16.getChar() is being called in new public method StringUTF16.compareTo(byte[], byte[], int, int) without appropriate boundary check. In the "old" code the "index" is guaranteed to be within [0, len) in StringUTF16.compareTo(byte[], byte[]), so it's safe. A real case for such scenario can be found in
JDK-8158168 [1], for example.

Thanks for the pointer! The email thread helps a lot. I've updated the webrev with a boundary check in ASB (AbstractStringBuilder line 106, 107), and then a note to StringUTF16.compareTo (StringUTF16 line 280). Hopefully this is sufficient. Didn't want to add any check in StringUTF16 since that may affect the original two-arg method.

JBS: https://bugs.openjdk.java.net/browse/JDK-8137326
webrev: http://cr.openjdk.java.net/~joehw/jdk11/8137326/webrev/



[1] https://bugs.openjdk.java.net/browse/JDK-8158168

Reply via email to