[
https://issues.apache.org/jira/browse/TIKA-1757?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14938914#comment-14938914
]
Yaniv Kunda commented on TIKA-1757:
-----------------------------------
Also, regarding the badness of {{URL#getFile()}} - on Windows machines it
returns a String starting with a slash - e.g. {{/C:\File.txt}}.
This, for some reason, when passed to a {{File}} constructor, is handled in a
lenient manner, and the preceding slash disappears - unlike
{{Paths.get(String)}} fails with a {{InvalidPathException}}.
> tika-batch tests fail on systems with whitespace or special chars in folder
> name
> --------------------------------------------------------------------------------
>
> Key: TIKA-1757
> URL: https://issues.apache.org/jira/browse/TIKA-1757
> Project: Tika
> Issue Type: Bug
> Reporter: Uwe Schindler
> Assignee: Tim Allison
> Attachments: TIKA-1757.patch
>
>
> This is one problem that forbiddenapis des not catch, because the method
> affected has valid use cases: {{URL#getFile()}} or {{URL#getPath()}} both
> return the URL path, which should never be treated as a file system path (for
> file: URLs). This is breaks asap, if the path contains special characters
> which may not be part of URL. getFile() and getPath() return the encoded path.
> The correct way to transform a file URL to a file is: {{new
> File(url.toURI())}}. See also the list of "bad stuff" as listed by the Maven
> community for Mojos/Plugins.
> In fact the affected test should not use a file at all. Instead it should use
> {{Class#getResourceAsStream()}}.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
