A recently reported bug shows a race condition is possible on the rem == 0 check in ZipFile.read(byte b[], int off, int len). A bad check can result in referencing a jzentry structure that might already be freed and hence result in a SEGV. The fix is trivial and involves moving the rem == 0 check into a synchronized block. The ZipFile API itself is not thread safe so having mutiple threads operate on the same ZipFileInputStream is something that should never be performed.

webrev : http://cr.openjdk.java.net/~coffeys/webrev.8038491/webrev/

regards,
Sean.

Reply via email to