Unable to decompress compressed tar files with ArchiveFactory.OpenArchive

[doxxx]

Using SharpCompress 0.47.4, when trying to decompress a gnu tar file compressed using gzip, bzip2 or zstd, I get the following error after the second entry:

dir  size=0 compressed=0 path=./
file size=4149673 compressed=4149673 path=./GHT00710.jpg
SharpCompress.Common.IncompleteArchiveException: Failed to read TAR header
   at SharpCompress.Archives.Tar.TarArchive.LoadEntries(IEnumerable`1 volumes)+MoveNext() in D:\a\sharpcompress\sharpcompress\src\SharpCompress\Archives\Tar\TarArchive.cs:line 179
   at SharpCompress.LazyReadOnlyCollection`1.LazyLoader.MoveNext() in D:\a\sharpcompress\sharpcompress\src\SharpCompress\LazyReadOnlyCollection.cs:line 53

If I use the same test code against an uncompressed tar file, it works fine.

Test code:

using var stream = File.OpenRead(archivePath);
using var archive = ArchiveFactory.OpenArchive(stream);

foreach (var entry in archive.Entries)
{
    var entryType = entry.IsDirectory ? "dir " : "file";
    var key = string.IsNullOrWhiteSpace(entry.Key) ? "<no-name>" : entry.Key;

    Console.WriteLine(
        $"{entryType} size={entry.Size} compressed={entry.CompressedSize} path={key}");
}

[doxxx]

If I use ReaderFactory instead, it works fine:

using var stream = File.OpenRead(archivePath);
using var reader = ReaderFactory.OpenReader(stream);

while (reader.MoveToNextEntry())
{
    var entry = reader.Entry;
    var entryType = entry.IsDirectory ? "dir " : "file";
    var key = string.IsNullOrWhiteSpace(entry.Key) ? "<no-name>" : entry.Key;

    Console.WriteLine($"{entryType} size={entry.Size} compressed={entry.CompressedSize} path={key}");
}

[adamhathcock]

I'm unsure about supporting this situation as it essentially one compressed file with a tar inside. Reading the entries without decompressing isn't possible

[doxxx]

I see your point. If you don't want to support IArchive for compressed tar files, I can find a different way to do what I need. It would most likely involve decompressing the tar to a temp file and then opening the tar file using ArchiveFactory. I need to interactively present a directory tree to the user so they can select a sub-tree to extract into the destination folder.

[adamhathcock]

I'm just thinking why the Reader wouldn't work for this beyond running the decompression twice?

I guess you want random access for the tar after decompressing once.

I'm not sure what is best at a high level 🤔

[doxxx]

I have two main use-cases:

1. Extracting the entire contents of an archive to a directory. This can be done using the Reader API.
2. Presenting a tree of the directories in an archive, from which the user can select a directory to extract. This is primarily what I need IArchive for, since I need to extract a list of directories from the archive before actually extacting the files themselves.

The archives in use-case 1 are generally large, but in use-case 2 they are usually quite small -- patches to be applied to a folder. I could just extract them to a temp dir using Reader, let the user pick the folder they want, copy the files across to the actual destination folder and then delete the temp folder.

[doxxx]

I also don't have any control over the archive format for either use-case -- they could be zip, rar, tar.bz2, 7z, etc.

[doxxx]

What also might be useful is a utility to detect whether a particular archive can quickly extract a full directory/file listing without decompressing the entire archive. For example, the zip format. That way I could have two code paths, one which uses IArchive and provides nice progress feedback, and one which uses IReader and only shows a spinner (since IReader's IProgress feedback is only per-entry extraction progress and not overall archive extraction progress).

[adamhathcock]

The last thing is an idea recorded here: #1298