[QUESTION] tar.x does not save directory structure.

[swateek]

I have a tar file with the structure below:

    mysample.tar.gz
                HEADER.txt directory1 directory2 directory3

   directory1/
          resolv.conf resov.conf1 a1.txt
     
   directory2/
          stats.txt

Upon using the below line

await tar.x({cwd: extractionPath, file: fullFilePath});

I see that all files lose their directories, and come under one folder mentioned as "cwd" above. I really need to maintain the folder structure.

tar xf mysample.tar.gz -C extractionPath which is supposed to behave similarly as tar.x doesn't match. I am not sure what else I should try.

[isaacs]

Works for me? Can you share the script you're actually running, and the tar tvf file.tgz output?

$ tree
.
├── dest
├── src.tgz
└── unpack.js

$ tar tvf src.tgz
-rw-r--r--  0 isaacs staff       0 Oct  1 13:32 HEADER.txt
drwxr-xr-x  0 isaacs staff       0 Oct  1 13:33 directory1/
-rw-r--r--  0 isaacs staff       0 Oct  1 13:33 directory1/a1.txt
-rw-r--r--  0 isaacs staff       0 Oct  1 13:33 directory1/resov.conf1
-rw-r--r--  0 isaacs staff       0 Oct  1 13:33 directory1/resolv.conf
drwxr-xr-x  0 isaacs staff       0 Oct  1 13:33 directory2/
drwxr-xr-x  0 isaacs staff       0 Oct  1 13:33 directory3/

$ cat unpack.js
const tar = require('tar')
const { resolve } = require('path')
const file = resolve('src.tgz')
const cwd = resolve('dest')
tar.x({
  file,
  cwd,
}).then(() => console.log('done'))

$ node unpack.js
done

$ tree
.
├── dest
│  ├── directory1
│  │  ├── a1.txt
│  │  ├── resolv.conf
│  │  └── resov.conf1
│  ├── directory2
│  ├── directory3
│  └── HEADER.txt
├── src.tgz
└── unpack.js

[alicewriteswrongs]

I don't know for sure if this is the case, but I just ran in to the same issue and the culprit was the strip option. I copy-pasted the example from the README.md for .x which looks like this:

fs.createReadStream('my-tarball.tgz').pipe(
  tar.x({
    strip: 1,
    C: 'some-dir' // alias for cwd:'some-dir', also ok
  })
)

I know I didn't read closely enough - perhaps that happened to @swateek as well?

[aminya]

tar strip doesn't work properly. See this:
iwatakeshi/gitly#10

[isaacs]

Ah, I see. This is a bug when there is strip: 1 used, and the archive contains files at the top level. In that case, tar thinks it has a file named '', which it correctly refuses to extract, but it makes it think that '' is not a directory, so subsequent unpacks are ignored as well.

This patch fixes it, will be addressed in next patch release

diff --git a/lib/unpack.js b/lib/unpack.js
index edaf783..6433bcb 100644
--- a/lib/unpack.js
+++ b/lib/unpack.js
@@ -205,6 +205,8 @@ class Unpack extends Parser {
       if (parts.length < this.strip)
         return false
       entry.path = parts.slice(this.strip).join('/')
+      if (entry.path === '' && entry.type !== 'Directory' && entry.type !== 'GNUDumpDir')
+        return false
 
       if (entry.type === 'Link') {
         const linkparts = entry.linkpath.split(/\/|\\/)

[isaacs]

Script demonstrating the error:

#!/bin/bash

rm -rf gh-261
mkdir gh-261
cd gh-261

mkdir -p directory{1,2,3}
touch HEADER.txt directory1/{a,b,c}

tree .

tar cvf gh-261.tar *
tar tvf gh-261.tar
rm -rf directory* HEADER.txt
tar tvf gh-261.tar

node - <<JS
const tar = require('../')
tar.x({
  file: 'gh-261.tar',
  strip: 1,
  onentry: entry => console.log(entry.path),
}).then(() => console.log('done'))
JS

tree .

rm -rf a b c

tar xvf gh-261.tar --strip=1

tree .