adding some bad parquet files

[jp0317]

Hi, these two bad files can help reproduce the two examples mentioned in apache/arrow-rs#6228.

[mapleFU]

Just wonder how these file is generated? 🤔 These files are little, which is good for our testing

[mapleFU]

@jp0317 I've check the bad-dict file, it throw here in C++ code:

  if (col_start < 0 || col_length < 0) {
    throw ParquetException("Invalid column metadata (corrupt file?)");
  }

  if (AddWithOverflow(col_start, col_length, &col_end) || col_end > source_size) {
    throw ParquetException("Invalid column metadata (corrupt file?)");
  }

So maybe arrow-rs can also check this? ( also cc @alamb )

[mapleFU]

@pitrou Would you mind check the naming of the 2 new bad file here?

Besides, the bad-levels file is similiar to existing bad level file, but I think one more bad file is ok. Want to hear what do you think of this

[alamb]

So maybe arrow-rs can also check this? ( also cc @alamb )

I agree we should add tests to parquet-rs to test these bad files. I'll see what i can do

[alamb]

So maybe arrow-rs can also check this? ( also cc @alamb )

I made a PR to check these files with Rust: apache/arrow-rs#6262

[jp0317]

@jp0317 I've check the bad-dict file, it throw here in C++ code:

  if (col_start < 0 || col_length < 0) {
    throw ParquetException("Invalid column metadata (corrupt file?)");
  }

  if (AddWithOverflow(col_start, col_length, &col_end) || col_end > source_size) {
    throw ParquetException("Invalid column metadata (corrupt file?)");
  }

So maybe arrow-rs can also check this? ( also cc @alamb )

looks like the col start and length are checked against negative here but isn't checked against the file length.

Beside, it seems we'd better check data_len against file length instead of "with_capacity() then take() here

[mapleFU]

I've an issue here: apache/arrow-rs#6255 , glad to learn some rust code lol