[COMPRESS-249] TarArchiveInputStream does not properly read from underlying InputStream

• Type: Bug
• Status: Closed
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 1.6
• Fix Version/s: 1.7
• Component/s: Archivers
• Assignee: Unassigned
• Reporter: Ilja Pavkovic
• Created: 10/Dec/13 12:37
• Updated: 10/Dec/13 13:13
• Resolved: 10/Dec/13 12:39

Description

TarArchiveInputStream reads header with its protected function readRecord(). This code improperly assumes that is.read(buf) always fills the whole buffer. This assumption is wrong. A proper implementation needs to call read multiple times until all bytes are read:

TarArchiveInputStream.java

  protected byte[] readRecord() throws IOException {
        byte[] record = new byte[recordSize];

        // start change
        // int readNow = is.read(buf);
        int readNow = 0;
        while(readNow < recordSize){
          int bytesRead = is.read(record,0, recordSize - readNow);
          if(bytesRead == -1) {
            break;
          }
          readNow += bytesRead;
        }
        // end change
        count(readNow);
        if (readNow != recordSize) {
            return null;
        }

        return record;
    }

Issue Links

• duplicates: COMPRESS-245 TarArchiveInputStream#getNextTarEntry returns null prematurely
• is broken by: COMPRESS-234 Patch: TAR InputStream Huge Speed Improvements