TAR: Implement extraction and archiving of hardlinks. #286

wilx · 2023-05-25T19:22:53Z

This implements hardlinks in TAR. There is also a test to prove it.

wilx · 2023-05-25T19:26:22Z

My intention with this is to be able to create TAR archives with hardlinks with maven-assembly-plugin.

However, I am not sure if this is the best possible way to do this. I am not sure if it fits right into the PlexusIoResource framework.

plamentotev

Thanks for contributing. Hard links is something we still missing to better support tar files.

This change does not cover some edge cases. Lets say file and link are file and hard link in a tar file. In the most straight forward case we are going to extract both files and everything will work fine. But there are other cases.

We can extract only link. If file is not present then I'm not sure the proper exception is caught in order to just create the file. But more importantly if file do exist on the file system, a hard link to it would be created instead of creating new file.

Plexus Archiver have a feature named file mappers. So after extracting file may be called someFile. What we should do in such cases?

Also does the order of the entries matter. Is it possible that link is extracted before file and in such case would file be hard link to link?

plamentotev · 2023-05-27T07:33:08Z

src/main/java/org/codehaus/plexus/archiver/tar/TarArchiver.java

            if (entry.getType() == ArchiveEntry.SYMLINK) {
-                final SymlinkDestinationSupplier plexusIoSymlinkResource =
-                        (SymlinkDestinationSupplier) entry.getResource();
+                final SymlinkDestinationSupplier plexusIoSymlinkResource = (SymlinkDestinationSupplier) ioResource;


Is this change related to hard links? It seems that it is just refactoring to avoid calling entry.getResource() multiple times. If this is the case I would rather have this in separate PR or at least separate commit as the change is already complex enough to understand on its own.

plamentotev · 2023-05-27T07:36:37Z

src/main/java/org/codehaus/plexus/archiver/tar/TarArchiver.java

+                final Path file = fileResource.getFile().toPath();
+                if (Files.exists(file)) {
+                    final BasicFileAttributeView fileAttributeView =
+                            Files.getFileAttributeView(file, BasicFileAttributeView.class);


I would rather use LinkOption.NOFOLLOW_LINKS to be on the safe side, although it seems that symbolic links are already handled (haven't tested it).

plamentotev · 2023-05-27T07:53:55Z

src/main/java/org/codehaus/plexus/archiver/tar/TarArchiver.java

                }
            }

+            boolean doCopy = true;


I found this name confusing. isLink, or something else, probably would be more appropriate as it seems that the flag indicates whether the entry is link (symbolic or hard) to some other entry.

plamentotev · 2023-05-27T07:55:47Z

src/main/java/org/codehaus/plexus/archiver/tar/TarArchiver.java

+                if (Files.exists(file)) {
+                    final BasicFileAttributeView fileAttributeView =
+                            Files.getFileAttributeView(file, BasicFileAttributeView.class);
+                    if (fileAttributeView != null) {


I think it would be best to extract this check to separate method to make this method easier to read.

plamentotev · 2023-05-27T07:56:38Z

src/main/java/org/codehaus/plexus/archiver/tar/TarArchiver.java

+                    }
+                }
+            }
+            if (te == null) {


To be honest I don't follow why this is added.

te is assigned either from the symlink branch or the hardlink branch. But the hardlink branch can also not assign value at all. This handles that case plus all other cases of non-symlink&non-hardlink.

plamentotev · 2023-05-27T08:16:16Z

src/main/java/org/codehaus/plexus/archiver/AbstractUnArchiver.java

            String entryName,
            final Date entryDate,
            final boolean isDirectory,
+            final boolean isSymlink,


This is breaking change. I wonder if we can avoid it.

I don't see how.

plamentotev · 2023-05-27T08:51:07Z

Also we need to be extra careful when extracting multiple copies instead of creating hard links. If a tar file contains single 1 MB file and thousands of hard links, then extracted (if files are copied) could be 1 GB, possibly resulting in DoS attack.

doCopy -> isLink

Use NOFOLLOW_LINKS to get file attributes for hardlink.

wilx · 2023-08-21T10:28:42Z

FYI, I will not be pursuing this pull request further. I was curious if it could be done. But it is unclear to me how to do this so that it matches all the requirements and design criteria of the library. Feel free to pick it up or close it.

Copilot

Pull request overview

This PR implements hardlink support for TAR archives, enabling both extraction and creation of hardlinks in TAR files. The implementation detects hardlinks during archiving by tracking file keys, creates appropriate TAR entries, and properly extracts them by creating actual hardlinks on the filesystem (with fallback to file copying when hardlink creation is unsupported).

Added hardlink detection during TAR archiving using file key tracking
Modified extraction logic to create hardlinks when encountered in TAR archives
Added comprehensive test coverage for the hardlink round-trip scenario

Reviewed changes

Copilot reviewed 8 out of 9 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
src/main/java/org/codehaus/plexus/archiver/tar/TarArchiver.java	Implements hardlink detection using file keys and creates LF_LINK entries; adds preserveHardLinks option to TarOptions
src/main/java/org/codehaus/plexus/archiver/tar/TarUnArchiver.java	Updates extraction to detect both symlinks and hardlinks, passing link destination to extractFile method
src/main/java/org/codehaus/plexus/archiver/AbstractUnArchiver.java	Adds hardlink creation logic with UnsupportedOperationException handling and copy fallback; adds isSymlink parameter to distinguish symlinks from hardlinks
src/main/java/org/codehaus/plexus/archiver/zip/AbstractZipUnArchiver.java	Updates extractFile call to include new isSymlink parameter
src/test/java/org/codehaus/plexus/archiver/HardlinkTest.java	Adds test verifying hardlink preservation through extraction and re-archiving
src/test/java/org/codehaus/plexus/archiver/tar/TarFileTest.java	Updates to skip hardlinks when iterating entries and improves assertion method
src/test/java/org/codehaus/plexus/archiver/AbstractUnArchiverTest.java	Updates extractFile method call to include new isSymlink parameter
src/test/resources/symlinks/regen.sh	Adds commands to generate hardlinks test archive
src/test/resources/hardlinks/hardlinks.tar	Binary TAR archive containing hardlink test data

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-01-01T13:31:46Z

src/main/java/org/codehaus/plexus/archiver/tar/TarArchiver.java


    private TarArchiveOutputStream tOut;

+    private final Map<Object, String> seenFiles = new HashMap<>(10);


The seenFiles map is never cleared between archive creation operations. If a TarArchiver instance is reused to create multiple archives, the map will contain stale file keys from previous archives, potentially causing incorrect hardlink references across different archive creation operations. Consider clearing the map at the beginning of the execute() method or in the cleanUp() method.

Copilot · 2026-01-01T13:31:47Z

src/main/java/org/codehaus/plexus/archiver/AbstractUnArchiver.java

+                    try {
+                        Files.createLink(
+                                targetFileName.toPath(),
+                                FileUtils.resolveFile(dir, linkDestination).toPath());
+                        doCopy = false;
+                    } catch (final UnsupportedOperationException ex) {
+                        if (warnCannotHardlink) {
+                            getLogger().warn("Creating hardlinks is not supported");
+                            warnCannotHardlink = false;
+                        }
+                        // We will do a copy instead.
+                    }


The hardlink creation code only catches UnsupportedOperationException but Files.createLink() can throw IOException for various error conditions (e.g., file already exists, insufficient permissions, I/O errors). These IOExceptions will propagate up and may not be handled appropriately, potentially causing the extraction to fail when it could fall back to copying the file content instead. Consider also catching IOException and falling back to copying in that case.

Copilot · 2026-01-01T13:31:47Z

src/main/java/org/codehaus/plexus/archiver/AbstractUnArchiver.java

+            boolean doCopy = true;
+            if (!StringUtils.isEmpty(linkDestination)) {
+                if (isSymlink) {
+                    SymlinkUtils.createSymbolicLink(targetFileName, new File(linkDestination));
+                    doCopy = false;
+                } else {
+                    try {
+                        Files.createLink(
+                                targetFileName.toPath(),
+                                FileUtils.resolveFile(dir, linkDestination).toPath());
+                        doCopy = false;
+                    } catch (final UnsupportedOperationException ex) {
+                        if (warnCannotHardlink) {
+                            getLogger().warn("Creating hardlinks is not supported");
+                            warnCannotHardlink = false;
+                        }
+                        // We will do a copy instead.
+                    }
+                }
            } else if (isDirectory) {
                targetFileName.mkdirs();
-            } else {
+                doCopy = false;
+            }


The hardlink creation logic has several issues:

Only UnsupportedOperationException is caught, but Files.createLink() can throw IOException for various error conditions (e.g., file already exists, insufficient permissions, I/O errors), which should also be caught to fall back to copying.

After creating links (hardlinks or symlinks), the code continues to execute outside this block and will call setLastModified() and chmod() on the link at lines 347 and 350. On most platforms, these operations on links affect the target file, not the link itself, which is likely not the intended behavior.

Consider catching IOException in addition to UnsupportedOperationException, and tracking whether a link was created to avoid calling setLastModified() and chmod() on links.

TAR: Implement extraction and archiving of hardlinks.

913783b

wilx force-pushed the master-hardlinks branch from 1ffb0f5 to 913783b Compare May 25, 2023 19:28

plamentotev suggested changes May 27, 2023

View reviewed changes

wilx added 3 commits May 27, 2023 20:33

TAR: Implement extraction and archiving of hardlinks. [2]

95c0f44

TAR: Implement extraction and archiving of hardlinks. [3]

dd5e14f

doCopy -> isLink

TAR: Implement extraction and archiving of hardlinks. [4]

cea72cc

Use NOFOLLOW_LINKS to get file attributes for hardlink.

slachiewicz requested a review from Copilot January 1, 2026 13:28

Copilot started reviewing on behalf of slachiewicz January 1, 2026 13:28 View session

Copilot AI reviewed Jan 1, 2026

View reviewed changes


		private TarArchiveOutputStream tOut;

		private final Map<Object, String> seenFiles = new HashMap<>(10);

TAR: Implement extraction and archiving of hardlinks. #286

Are you sure you want to change the base?

TAR: Implement extraction and archiving of hardlinks. #286

Uh oh!

Conversation

wilx commented May 25, 2023

Uh oh!

wilx commented May 25, 2023

Uh oh!

plamentotev left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

plamentotev commented May 27, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

wilx commented Aug 21, 2023

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Jan 1, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 1, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 1, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

plamentotev commented May 27, 2023 •

edited

Loading