Improve robustness when the wrong encoding was passed into ZipArchiveInputStream / ZipFile

If a zip file is in one encoding and I try to read it using a different encoding, what I expect to happen is that the filenames get garbled but the data otherwise extracts correctly (which is what I see using native tools to extract a zip file in this fashion.)

However, what Commons Compress can do is to try and decode a name, fail, and ultimately give us no zip entries to work with.

Here's a test to show what I mean:

@Test
public void testWrongEncodingPassedIn() throws Exception {
// Making the test zip file:
File inputFile = new File(scratch, "test.dat");
byte[] inputData =

{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9 }

;
FileUtils.writeByteArrayToFile(inputFile, inputData);
File file = new File(scratch, "test.zip");
try (ZipArchiveOutputStream out = new ZipArchiveOutputStream(file))

{ out.setEncoding("windows-31j"); ZipArchiveEntry entry = new ZipArchiveEntry(inputFile, "\u767A\u8D77\u4EBA\u6C7A\u5B9A\u66F8"); out.putArchiveEntry(entry); out.write(inputData); out.closeArchiveEntry(); }

// Trying to iterate over it:
int entryCount = 0;
try (ZipArchiveInputStream in = new ZipArchiveInputStream(new FileInputStream(file), "windows-1252", false)) {
ZipArchiveEntry entry = in.getNextZipEntry();
if (entry == null)

{ break; }

entryCount++;
}

assertEquals("Wrong number of entries", 1, entryCount);
}

In this situation it's definitely the caller's "fault", but unfortunately the end user is often the one supplying the encoding and they would rather see garbled contents with the actual data intact, than no data at all.