Partial gzip header breaks decoding

[jongiddy]

The read_gz_header_part function returns an Err if the first 10 bytes of the gzip header are not present or if the FEXTRA flag is set and the extra fields are not present. Returning an Err causes GzDecoder::write to keep the buffer and wait for more data to complete the header.

However, if the header contains an optional filename or comment but they are not completely contained in the buffer, then a valid header is created, missing data from the optional sections. The next write call will treat the remaining header as encoded data and attempt to unzip it.

The problem occurs because for byte in r.bytes() ends the iteration when the end of the buffer is found, falling through to the next state.

If this happens, the code should return Err(io::ErrorKind::UnexpectedEof) to allow a subsequent write to complete the header.

[jongiddy]

Anyone trying to fix this issue probably needs to look at the discussion here: https://github.com/rust-lang/flate2-rs/pull/278/files#r676719137

[jongiddy]

#265 is possibly linked - missing filename in header and problems decompressing.