Sorry, pressed "send" too soon. There's no information on how this fix
was tested. Right now we're seeing a varying number of failures in almost
every Tier5 CI job set. Please verify that you've tested Tier5.
Dan
On 4/9/20 5:01 PM, Daniel D. Daugherty wrote:
On 4/9/20 4:42 PM, Alex Menkov wrote:
Hi all,
Please review the fix for
https://bugs.openjdk.java.net/browse/JDK-8242282
webrev:
http://cr.openjdk.java.net/~amenkov/jdk15/jpsTest_ClsNotFound/webrev/
test/jdk/sun/tools/jps/LingeredAppForJps.java
L89: manifestClasspath += " " + new File(path
).toURI();
nit - please delete extra space before ')'.
old L91: break;
So the old version only added the first existing file found in
the
path to the jarArgs. The new version is added all existing files.
Maybe I misread the bug report, but I didn't think you wanted to
do that.
Also, can you provide an example of the old manifest file
versus the
new manifest file after this change?
test/jdk/sun/tools/jps/TestJps.java
No comments.
The test creates jar with test classes and run it with "java -jar
<jar_file>".
The problem is single "@run Test" tag is executed by JTreg
inconsistently - sometimes library classes are compiled to test
directory, sometimes - to library directory.
The fix explicitly forces required library classes compilation (so
library files are compiled to library dir) and add classpath
directories to jar manifest.
This description doesn't mention jar'ing up additional files so you
can see
my confusion (I hope).
Dan
--alex
