[jira] [Commented] (COMPRESS-469) java.io.IOException: Error detected parsing the header on un-archiving. Archive untars fine in BSDTar and JTar

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

    [ https://issues.apache.org/jira/browse/COMPRESS-469?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16681827#comment-16681827 ] 

Stefan Bodewig commented on COMPRESS-469:
-----------------------------------------

GNU tar says

{code}
$ tar tvvf assembly-0.2.12.tar
...
-rw-r--r-- 501/20          193 2012-02-29 18:15 package/test/fixtures/amd/expected.js
tar: Archive contains ‘00000000NaN’ where numeric time_t value expected
-rw-r--r-- 501/20           74 1970-01-01 00:59 package/test/fixtures/app.js
...
tar: Exiting with failure status due to previous errors
{code}

Extracting also yields

{code}
tar: package/test/fixtures/app.js: implausibly old time stamp 1970-01-01 00:59:59
{code}

so at least we are not alone :-)

There is no question the archive is broken but this doesn't help you.

The field in question is the timestamp of the entry. I'll have a look at what bsdtar uses as fallback, GNU tar seems to use -1 as 1970-01-01 00:59:59 would be a second before the Unix epoch translated to my timezone (UTC+1).

Can you tell which timestamp bsdtar creates for package/test/fixtures/app.js on your machine?


> java.io.IOException: Error detected parsing the header on un-archiving. Archive untars fine in BSDTar and JTar
> --------------------------------------------------------------------------------------------------------------
>
>                 Key: COMPRESS-469
>                 URL: https://issues.apache.org/jira/browse/COMPRESS-469
>             Project: Commons Compress
>          Issue Type: Bug
>    Affects Versions: 1.18
>         Environment: Java version: 1.8.0_192, vendor: Oracle Corporation, runtime: /Library/Java/JavaVirtualMachines/jdk1.8.0_192.jdk/Contents/Home/jre
>  Default locale: en_US, platform encoding: UTF-8
>  OS name: "mac os x", version: "10.14", arch: "x86_64", family: "mac"
>  
> bsdtar 2.8.3 - libarchive 2.8.3
>            Reporter: Kirk Hardy
>            Priority: Major
>         Attachments: compress-bug-POC.tar.gz
>
>
> Using TarArchiveInputStream to parse this tar archive results in a
> `java.io.IOException: Error detected parsing the header`
> `Caused by: java.lang.IllegalArgumentException: Invalid byte 78 at offset 8 in '00000000NaN\{NUL}' len=12`
>  
> The `00000000NaN\{NUL}` value is indeed found in the tar archive at the point that TarArchiveInputStream finds it and errors, but both BSDTar from a bash shell and JTar from a Java environment handle the archive fine, including the file with the malformed header. My commons-compress code is:
> {code:java}
> private static List<File> unTar(final File inputFile, final File outputDir) throws FileNotFoundException, IOException, ArchiveException {
>   System.out.println(String.format("Untaring %s to dir %s.", inputFile.getAbsolutePath(), outputDir.getAbsolutePath()));
>   final List<File> untaredFiles = new LinkedList<File>();
>   final InputStream is = new FileInputStream(inputFile);
>   final TarArchiveInputStream debInputStream = (TarArchiveInputStream) new ArchiveStreamFactory().createArchiveInputStream("tar", is);
>   TarArchiveEntry entry = null;
>     while ((entry = (TarArchiveEntry) debInputStream.getNextEntry()) != null) {
>       final File outputFile = new File(outputDir, entry.getName());
>       if (entry.isDirectory()) {
>         System.out.println(String.format("Attempting to write output directory %s.", outputFile.getAbsolutePath()));
>         if (!outputFile.exists()) {
>           System.out
>               .println(String.format("Attempting to create output directory %s.", outputFile.getAbsolutePath()));
>           if (!outputFile.mkdirs()) {
>             throw new IllegalStateException(
>                 String.format("Couldn't create directory %s.", outputFile.getAbsolutePath()));
>           }
>         }
>       }
>       else {
>         System.out.println(String.format("Creating output file %s.", outputFile.getAbsolutePath()));
>         final OutputStream outputFileStream = new FileOutputStream(outputFile);
>         IOUtils.copy(debInputStream, outputFileStream);
>         outputFileStream.close();
>       }
>       untaredFiles.add(outputFile);
>     }
>   debInputStream.close();
>   return untaredFiles;
> }
> {code}
> [^compress-bug-POC.tar.gz]
> I've attached an archive of my proof-of-concept code that un-archives the problem archive using JTar, then attempts to un-archive it using commons-compress. The problem archive is included. Unfortunately I wasn't involved in the creation of the problem archive, it was downloaded from NPM last year, so I don't have any details about the archival process that caused the malformed header.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)