On 15/08/2019 12:53, Andrew Dinn wrote:
Hi Alan,
Any further input on this patch forthcoming?
I think the main changes since I looked at it previously have been in
the tests.
On non-Linux platforms, FileChannel.map should throw UOE when invoked
with a mode map of READ_ONLY_SYNC or READ_WRITE_SYNC so I think we need
a test for that.
MapFail seems fragile. If you add @modules java.base/jdk.internal.misc
to the test description then it could use Unsafe::isWritebackEnabled and
I think would simplify the test and checks. It would mean it could run
on all platforms. Also "MapFail" is probably too general a name for a
test in MBB because its specific to the extended map modes, not a
general test of map failing.
In passing, MappedByteBuffer load/isLoaded check the fd value before
isSync, can force() do the same? Also the @return on the private isSync
method is very wordy and I don't think needs to duplicate the method
description.
-Alan
