Error in calculating ZipEntry._LengthOfTrailer when false DataDescriptor header found in compressed data

[DannyBoyk]

EPPlus usage

Commercial use (I have a commercial license)

Environment

Windows

Epplus version

7.5.2

Spreadsheet application

Excel

Description

When the ExcelPackage (zip) itself is using DataDescriptor headers, but a file in the package happens to contain the DataDescriptor header byte signature (0x08074b50) in the compressed bytes of the file itself, the logic in OfficeOpenXml.Package.Ionic.Zip.ZipEntry::ReadHeader is incorrectly incrementing the _LengthOfTrailer value by 16 (or 24 for Zip64 entries) for every iteration of the while (wantMore) loop. It should instead just be assigning the value of 16 (or 24 for Zip64 entries) to _LengthOfTrailer, not incrementing.

In my case, the wantMore loop was iterated over twice and the _LengthOfTrailer value was set to 32, when it should have been set to 16 for the single DataDescriptor header at the end of the compressed bytes. This caused it to skip the LocalEntry header of the next entry and fail to process the file with a BadReadException.

[DannyBoyk]

Here is a file that exhibits the issue. I generated it by putting a binary file in the folder structure and zipping on OSX with the -fd option to force data descriptors and -0 for no compression.
Book1.xlsx

[JanKallman]

Thanks for reporting this. I will have a look at it.

[JanKallman]

The attached file cannot be opened in Excel either.
Just so I understand. When is this a problem?

[DannyBoyk]

It can be opened in Excel, but it complains about the random binary file being in there and attempts to "repair" it. I will work to put together a sample that includes an image in the spreadsheet that I will put the DataDescriptor header bytes into. I can't put the original file that we have on our network here, so I was trying to put together something to replicate the zip parsing behavior.

The point of the file was just to cause the BadReadException for the zip processing. The change I indicated will fix that issue.

[JanKallman]

OK, thanks for the update. I will continue to look at this.

[DannyBoyk]

@JanKallman - please find attached a sample file that Excel has no problems opening, but EPPlus does. I embedded a JPEG image into the test spreadsheet that has the DataDescriptor header inside of it, which causes the problems described above.

I've confirmed that my fix allows EPPlus to parse the file.
From:

                        ze._LengthOfTrailer += 16;  // bytes including sig, CRC, Comp and Uncomp sizes

To:

                        ze._LengthOfTrailer = 16;  // bytes including sig, CRC, Comp and Uncomp sizes

And the same for the += 24 for Zip64.

Book1.xlsx

[JanKallman]

Thanks for your investigation and sample, I'll provide the fix.
It looks like the same issue as in this pr in DotNetZip.Semverd ...
https://github.com/haf/DotNetZip.Semverd/pull/110/files

[DannyBoyk]

If it's possible to get a release out for this soonish, that would be great.

[JanKallman]

Yes, this is out in EPPlus 7.5.3, released a moment ago.

[JanKallman]

I'll close this issue. Feel free to re-open it if you find any problems related to it.