[COMPRESS-6] [compress] TarOutputStream.java long file name bug (and fix!)

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed

Environment

Operating System: other
Platform: All

• Bugzilla Id: 30495
• Assignee: Torsten Curdt
• Reporter: Derc Yamasaki
• Created: 06/Aug/04 00:10
• Updated: 02/May/23 17:11
• Resolved: 07/Jan/09 12:54

Description

'LongLink' entries written to a tar file header by
TarOutputStream.putNextEntry() specify the wrong file name length and data,
preventing GNU tar from properly reading long file name entries (in the current
GNU 'tar' implementation, sometimes one gets lucky and the entry happens to be
null-terminated so it works anyway). In particular, the file name
length-stored as an octal string in each header entry-should include an extra
byte for the null terminator that must also be written at the end of the file
name in the header.

Here's what the code in
org.apache.commons.compress.tar.TarOutputStream.putNextEntry() currently looks
like (starting around line 424):

if( m_longFileMode == LONGFILE_GNU ) {
// create a TarEntry for the LongLink, the contents
// of which are the entry's name
final TarEntry longLinkEntry =
new TarEntry( TarConstants.GNU_LONGLINK,
TarConstants.LF_GNUTYPE_LONGNAME );

longLinkEntry.setSize( entry.getName().length() );
putNextEntry( longLinkEntry );
write( entry.getName().getBytes() );
//write( 0 );
closeEntry();
}

Here's what the code should have been:

if( m_longFileMode == LONGFILE_GNU ) {
// create a TarEntry for the LongLink, the contents
// of which are the entry's name
final TarEntry longLinkEntry =
new TarEntry( TarConstants.GNU_LONGLINK,
TarConstants.LF_GNUTYPE_LONGNAME );

longLinkEntry.setSize( entry.getName().length() + 1 );
putNextEntry( longLinkEntry );
write( entry.getName().getBytes() );
write( 0 );
closeEntry();
}