[jira] [Commented] (COMPRESS-330) Tar UnArchive Fails when archive contains directory sizes which are non-zero.

["Dye357 (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/COMPRESS-330?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15132280#comment-15132280 ] 

Dye357 commented on COMPRESS-330:
---------------------------------

I can confirm the tar archive was created by 7zip, still trying to figure out which version. (The file was sent to me by another party)

I debugged the tar in question this morning, I discovered the entrySize for the directory which causes the error is coming back as 32768. I thought that was a little suspicious since that is the max negative size of an int in various languages. So I quick wrote a unit test just to understand how the parseOctalOrBinary Method worked:

    @Test
    public void testParseOctal() {
        byte[] buffer = new byte[12];
        buffer[0] = 32;
        buffer[1] = 32;
        buffer[2] = 32;
        buffer[3] = 32;
        buffer[4] = 32;
        buffer[5] = 49;
        buffer[6] = 48;
        buffer[7] = 48;
        buffer[8] = 48;
        buffer[9] = 48;
        buffer[10] = 48;
        buffer[11] = 32;

        long result = TarUtils.parseOctalOrBinary(buffer, 0, 12);
        assert result == 32768;
    }

The byte array above is an exact copy of the entry size header of the file in question - I tried setting the entry size in my dummy tar to that exact byte array but my unit test still passes. When I use the file in question I get an IllegalArugmentEcxeption: Invalid byte 85 at offset 0 in '<randomBinaryString>' len=12

> Tar UnArchive Fails when archive contains directory sizes which are non-zero.
> -----------------------------------------------------------------------------
>
>                 Key: COMPRESS-330
>                 URL: https://issues.apache.org/jira/browse/COMPRESS-330
>             Project: Commons Compress
>          Issue Type: Bug
>          Components: Archivers
>    Affects Versions: 1.10
>         Environment: Java 1.8
>            Reporter: Dye357
>            Priority: Minor
>         Attachments: Compress-330.patch
>
>
> Tar UnArchive Fails when archive contains directory sizes which are non-zero. I recently came across a set of files which failed to extract with commons-compress but I was able to successfully extract the files with GNU tar. 
> The problem is TarArchiveInputStream.java gets the size of each entry in a tar archive to determine how many bytes to read ahead. Directories are always sized 0 bytes, however its technically possible a tar archive contains a size for a directory. This causes the Input Stream to loose it's place and eventually results in an exception being thrown.
> I was able to implement a proof-of-concept fix by checking if an entity is a directory and setting the directory size to 0. (A directory with Files in a Tar Archive still has a size of 0 as the directory itself does not have size in the archive.)



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)