[
https://issues.apache.org/jira/browse/HDFS-4632?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13612750#comment-13612750
]
Chris Nauroth commented on HDFS-4632:
-------------------------------------
This problem is visible by running {{TestGlobPaths#pTestEscape}} on Windows.
On Windows, the backslash will get converted to forward slash before
{{GlobExpander}} interprets it. This problem was not caught on branch-1-win,
because the test is commented out on that branch with a comment pointing to
HADOOP-1995. That jira contains a lot of discussion on this topic, so it would
be good for anyone participating on this jira to review that one first.
This is particularly awkward, because {{FileSystem#globStatus}} is a public API
accepting a {{Path}}. We can't change the normalization behavior of {{Path}}
without breaking a lot of things. We can't change the API to accept a
different kind of object (i.e. a new RawGlobPath class) without breaking
backwards compatibility. I can think of a few options:
# Accept this as a known limitation on Windows and skip running the failing
test on Windows. This is the conclusion of HADOOP-1995, so refer to that jira
for the full justification. This may in fact be the only viable short-term
solution.
# Change {{Path}} to maintain a separate raw, unescaped copy of the path at
construction time. Change {{FileSystem#globStatus}} to use this raw copy of
the path and allow all other clients of {{Path}} to continue using the escaped
version. The negative consequence is that this would effectively double the
memory consumption of {{Path}} instances.
# Add a new overload of {{FileSystem#globStatus}} that accepts something
different from {{Path}}. Deprecate the old one and gradually phase it out.
> globStatus using backslash for escaping does not work on Windows
> ----------------------------------------------------------------
>
> Key: HDFS-4632
> URL: https://issues.apache.org/jira/browse/HDFS-4632
> Project: Hadoop HDFS
> Issue Type: Bug
> Affects Versions: 3.0.0
> Reporter: Chris Nauroth
> Assignee: Chris Nauroth
>
> {{Path}} normalizes backslashes to forward slashes on Windows. Later, when
> passed to {{FileSystem#globStatus}}, the path is no longer treated as an
> escape sequence.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
