On 6/21/13 1:03 PM, Alan Bateman wrote:
Naoto,
As the test might copy a currency.data into the JDK under test then it
makes me wonder if this might cause interference for tests that are
running concurrently (in other VMs). We might have to adding this
directory to the exclusiveAccess.dir list (I don't know if there is an
equivalent option that can be added to the @run tag).
Good point. I will fix this later as it is not inherently related to
this bug.
Otherwise, I'm scratching my head a bit on why the changes are needed.
Clearly it fixes the case where $TESTJAVA is writable but the lib or
jre/lib directory is not.
Apparently the bug description claims the situation. That's why I moved
the writable check from the top dir to the exact dir where the test
copies the properties file.
Or is the main fix the cygpath -u ${TESTJAVA}
so that TESTJAVA has the right path for -w ?
This is merely working around a bug with Cygwin's writable check (`test
-w d:/foo` returns 1 even though the directory is not writable, while
/cygdrive/d/foo returns 0)
Naoto
If so then the changes look
fine to me.
-Alan.
On 21/06/2013 18:30, Naoto Sato wrote:
Still need a reviewer for this one.
Naoto
On 6/18/13 1:35 PM, Naoto Sato wrote:
Hello,
Please review this test case fix:
http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=6863624
http://cr.openjdk.java.net/~naoto/6863624/webrev.00/
Basically it enforces the check for non-writable JDKs, with a workaround
for Cygwin's inconsistent behavior for `test` command. Changes for
LocaleProviders are irrelevant for this fix, just the left over from
previous fix for 8015960.
Naoto
