Good catch! The merge was indeed quite hairy. Fixed here:
http://cr.openjdk.java.net/~erikj/8009280/webrev.05/
/Erik
On 2013-10-22 18:42, Magnus Ihse Bursie wrote:
There seem to have been some kind of merge error (?) at the end of
CompileJavaClasses.gmk; the all target is duplicated three times with
variations.
Apart from this, it looks good from a build perspective.
/Magnus
22 okt 2013 kl. 16:52 skrev Erik Joelsson <erik.joels...@oracle.com
<mailto:erik.joels...@oracle.com>>:
Trying this again. Here is a new webrev against jdk8/build. Again,
this solves the issue detailed by Alan in this bug.
http://cr.openjdk.java.net/~erikj/8009280/webrev.04/
<http://cr.openjdk.java.net/%7Eerikj/8009280/webrev.04/>
There will probably be more issues in "JDK-8006350 [build] JCE
unusable using new build environment", but at least this part has had
a fix ready for months. It would be nice to finally get it resolved.
If the JCE team is unable to verify this, perhaps a review from
someone else will be sufficient?
/Erik
On 2013-09-30 15:01, Erik Joelsson wrote:
Reposting an updated webrev for this:
http://cr.openjdk.java.net/~erikj/8009280/webrev.03/
<http://cr.openjdk.java.net/%7Eerikj/8009280/webrev.03/>
From what I can tell, this changes the jdk (default) target to
create a valid jce layout. The jdk_security* tests are not failing
for me. There will likely be more issues to resolve for the JCE
team, but this should cover what's reported in this particular bug.
/Erik
On 2013-08-16 11:35, Erik Joelsson wrote:
Woops, as some of you pointed out, I missed adding the link to the
webrev:
http://cr.openjdk.java.net/~erikj/8009280/webrev.02/
<http://cr.openjdk.java.net/%7Eerikj/8009280/webrev.02/>
/Erik
On 2013-08-16 10:46, Erik Joelsson wrote:
Finally able to revisit this patch:
With this patch the security tests will again be runnable on the
exploded jdk image. The main changes are:
* The security classes are compiled separately to a different
output directory.
* The security jars are created in the jdk target (instead of
images) and put in the jdk/lib/... directories.
Also did:
* Removed now redundant entries in rt.jar exclude list
* Changed source location for signing unsigned jars
* Made the SetupJavaCompilation macro more friendly with multiple
setups sharing output directories.
The reason it had to wait was that this design did not work with
sjavac. Sjavac has now been fixed in jdk8-tl.
/Erik