Fix long names (GNUTYPE_LONGNAME)

[frederikbosch]

The patch suggested by @RazerM in #369 is fixing my problem of a truncated file name during extraction. In my case the long file names ended up in the root of the extraction directory rather than the sub-directory they have in the archive.

I'm also not sure what the best way forward is for handling this is. I can't find anything about why these bytes might be different than expected, if not just a mixup of \0 and 0.

Is what @RazerM said last week. I'll be honest here: I don't have a clue at all. I'm just creating a PR for the solution that is solving my problem.

[RazerM]

To be clear, the issue is not related to long links.

P.S. I did not submit this as a PR myself because it should probably not be part of is_ustar – the header in the problematic tar files is non-standard.

[frederikbosch]

Are you sure about that @RazerM? In my case, I am creating archives in exactly the same manner. The only difference is the length of the added paths. I had a look in my archiving library that I created a long time ago. It contains a specific ././@LongLink case for files with dirnames longer than 155 bytes or filenames longer than 100 bytes. And your code is a solution.

[RazerM]

The long links are stored in a header whose version bytes are "wrong", tar-rs does not look in invalid headers.

I am creating archives

Using what? Sounds like it's creating out of spec headers.

[frederikbosch]

Apparently, I am using GNUTYPE_LONGNAME, see the link to my archive generation code in the message above, which is supported by Python.

[RazerM]

I mean what program or library are you using to create the archive

[frederikbosch]

See this library, the link references the specific line that is using the GNUTYPE_LONGNAME (L type header).

[RazerM]

I glossed over the link the first time because it's called TarReader. In any case I'm not going to review it. The library should be creating a valid header with the correct magic bytes and version.

[frederikbosch]

Maybe you are right, this is not about longname.

[frederikbosch]

Thanks for waking me up. I have fixed the ustar version issue in my own library. Much appreciated.