[jira] [Created] (COMPRESS-337) TAR header parsing attempts OLDGNU format on POSIX/STAR header

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

Jeremy Gustie created COMPRESS-337:
--------------------------------------

             Summary: TAR header parsing attempts OLDGNU format on POSIX/STAR header
                 Key: COMPRESS-337
                 URL: https://issues.apache.org/jira/browse/COMPRESS-337
             Project: Commons Compress
          Issue Type: Bug
          Components: Archivers
    Affects Versions: 1.10
            Reporter: Jeremy Gustie


This is at least tangentially related to COMPRESS-336, we found an archive which is misinterpreted as an OLDGNU header format when it appears to some type of STAR header (unlike COMPRESS-336 this one uses a NULL delimiter instead of space terminators for the time stamps). Apparently the archive itself comes from an [EyeFi card|https://github.com/golang/go/issues/5290](?!).

Both the original archive ([available here|https://storage.googleapis.com/go-attachment/5290/0/in.tar]) and the (presumably hand made) version for testing ([here|https://github.com/golang/go/raw/master/src/archive/tar/testdata/nil-uid.tar]) exhibit the same behavior: they fail with an {{IllegalArgumentException}} attempting to parse the old GNU {{realSize}} from a buffer partially filled with the {{atime}}/{{ctime}} values.

At first I thought if the {{TarArchiveEntry.evaluateType}} were to also look at the {{version}} field for {{VERSION_GNU_SPACE}} before returning {{FORMAT_OLDGNU}} it _should_ be enough (according to something I saw in [the GNU documentation|http://www.gnu.org/software/tar/manual/tar.html#SEC184] which defines {{OLDGNU_MAGIC "ustar  "  /* 7 chars and a null */}}). Unfortunately it looks like that is the magic/version used by the archives in question, leaving me stumped as to how to identify which of the three prevailing header formats needs to be parsed.



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