On Fri, 19 Nov 2021 10:10:33 GMT, Andrew Leonard <aleon...@openjdk.org> wrote:
>> Both jar and jmod utilise java.io file operations whose methods define no
>> ordering of the return file lists, and in fact rely on OS query file
>> ordering, which can differ by underlying OS architecture.
>> This PR adds sort processing to the creation of such jar's and jmod's to
>> enable a deterministic content ordering.
>>
>> Signed-off-by: Andrew Leonard <anleo...@redhat.com>
>
> Andrew Leonard has updated the pull request incrementally with one additional
> commit since the last revision:
>
> 8276764: Enable deterministic file content ordering for Jar and Jmod
>
> Signed-off-by: Andrew Leonard <anleo...@redhat.com>
The source change looks good. I have a couple comments for the tests.
test/jdk/tools/jar/ContentOrder.java line 55:
> 53: import jdk.test.lib.util.FileUtils;
> 54:
> 55: public class ContentOrder {
It'd be useful to include the test cases to verify the `--release` option with
the specified files and directories as well.
test/jdk/tools/jar/ContentOrder.java line 89:
> 87: jar("cf test.jar testjar");
> 88: jar("tf test.jar");
> 89: println();
Suggestion:
System.out.println(new String(baos.toByteArray()));
You can simply print the output from `jar tf` command here and the `println`
method isn't really needed.
test/jdk/tools/jar/CreateJarBenchmark.java line 56:
> 54: import jdk.test.lib.util.FileUtils;
> 55:
> 56: public class CreateJarBenchmark {
This is a benchmark which should not be run as regression testing. I wonder
if we need this benchmark for future use. If so I would suggest to convert it
to JMH and place it under `test/micro/org/openjdk/bench/tools/jar` directory.
I'm okay with or without it in this patch.
-------------
PR: https://git.openjdk.java.net/jdk/pull/6395
