[TIKA-3800] Consider wrapping 'unrar' commandline executable as a parser to handle rar v5

• Type: Task
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Fix Version/s: 2.5.0
• Assignee: Unassigned
• Reporter: Tim Allison
• Created: 23/Jun/22 19:11
• Updated: 26/Sep/23 16:29
• Resolved: 24/Jun/22 21:44

Description

Junrar is great and doesn't require any external dependencies. However, it doesn't handle rar v5. I've tried UNRAR 5.61 beta 1 freeware on some of the v5 files that we have in our regression corpus, and I can confirm that Tika is not able to handle them, but unrar is.

The parser would need to create a temporary directory, copy the inputstream there to a file, run unrar, process the extracted files and then clean up the directory.

We can get full path information from the l command: unrar l blah.rar

We can tell unrar not to overwrite files with the same name: unrar e or bug_trackers/LIBRE_OFFICE/131138-137877/LIBRE_OFFICE-135119-0.rar.

If we trust unrar to protect against path traversal (e.g. an embedded file with the name "../../../something_bad.pdf"), we can use the x command.