RFR : 7148584 Jar tools fails to generate manifest correctly when boundary condition hit (original) (raw)

Seán Coffey sean.coffey at oracle.com
Mon Mar 12 09:28:07 UTC 2012

Alan,

I did (briefly) look into using the java.util.jar API but wasn't seeing the bug being triggered. Decided to use the jar process command as an alternative.

The testcase was a port from JDK 6 and didn't contain JDK 7+ features. I've converted it now though. Hope it's cleaner. http://cr.openjdk.java.net/~coffeys/webrev.7148584.jdk8.2/

regards, Sean.

On 09/03/2012 16:18, Alan Bateman wrote:

On 09/03/2012 15:52, Seán Coffey wrote:

Issue seen when the inner Manifest.FastInputStream.peek() method is called just as we hit EOF of main buffer being parsed (the manifest input file)

Simple fix involves getting peek() to return -1 if a fill() request fails to read anything. webrev : http://cr.openjdk.java.net/~coffeys/webrev.7148584.jdk8/ Bug report : http://bugs.sun.com/bugdatabase/viewbug.do?bugid=7148584 regards, Sean. The change looks okay to me. On the test case then I assume it could have been done with the java.util.jar APIs rather than calling into the jar command. Also it might be cleaner to use try-with-resources. -Alan.

More information about the core-libs-dev mailing list