C# ZipArchive Illegal characters in path

Question

Given the following code:

using (var data = new MemoryStream(bytes))
using(var archive = new ZipArchive(data))
{
    foreach (var entry in archive.Entries)
    {
        entry.FullName.Log();
    }
    ...

The exception thrown is:

Illegal characters in path.

On the foreach line.

How do I work out which the affected entry is? Whenever I try to access the entries it throws the exception. It appears that this specific archive is created from a Mac as it contains the _MACOSX folder.

The stack trace:

[ArgumentException: Illegal characters in path.]
   System.IO.Path.CheckInvalidPathChars(String path, Boolean checkAdditional) +14351233
   System.IO.Path.GetFileName(String path) +29
   System.IO.Compression.ZipHelper.EndsWithDirChar(String test) +9
   System.IO.Compression.ZipArchiveEntry.set_FullName(String value) +93
   System.IO.Compression.ZipArchiveEntry..ctor(ZipArchive archive, ZipCentralDirectoryFileHeader cd) +228
   System.IO.Compression.ZipArchive.ReadCentralDirectory() +172
   System.IO.Compression.ZipArchive.get_Entries() +36

Answer 1

Your exception is being thrown because ZipArchive.Entries attempts to read the zip central directory in order to construct a ZipArchiveEntry for each entry. The ZipArchiveEntry constructor is calling Path.GetFileName, which calls Path.CheckInvalidPathChars, which throws if the path contains \0.

Although an exception part-way through reading the central directory will leave the ZipArchive with a partially-populated list of entries, there doesn't seem to be any way to read it without triggering a read of the central directory again, unless you use the debugger / reflection to read _entries or _entriesCollection. This will tell you which entries succeeded, which might narrow down the failing one a bit. If you're going to these lengths, debugging into the .Entries call would tell you exactly which entry is failing.

It looks like a related issue with paths being treated differently across different platforms was recognised in 2015, and as part of that a change was made to get the file name without calling Path.CheckInvalidPathChars, which should avoid the issue (you'll still get an exception if you try to write this entry to a file on Windows, but at least .Entries shouldn't throw). Unfortuatenly this was only introduced in .NET Core however.

Answer 2

Agree with comments. Are you sure your file is intact and originates from a compatible platform.. What happens if you open your Archive with e.g. WinRar ?

I tested below code before and successfully modified it using a memory stream like you do, and I am sure your version should work as well..

    [Test]
    public void TestZipper()
    {
        using (var dataFromFile = new FileStream("d:\\lx\\MyZip.ZIP", FileMode.Open))
        {
            var dataInMemory = new MemoryStream();
            dataFromFile.CopyTo(dataInMemory);
            using (var archive = new ZipArchive(dataInMemory))
            {
                foreach (var entry in archive.Entries)
                {
                    Debug.WriteLine(entry.FullName);
                }
                Assert.That(MyZipListIsComplete(archive.Entries));
            }
        }
    }