Those attacks are DoS-only.

- Rework the testsuite wrt. torrent
- fail at parser's instantiation on corrupted torrent,
  instead of during `get_meta` or `remove_all` call

We're using methods that aren't available in Poppler
below 0.46, so we're checking for this upon import.

This commit is based on ideas from @LogicalDash 

- xml files are no longer considered harmless
- Factorization of the `remove_all` method for office files
- Explicit whitelist are used
- Blacklist are used to skip files completely
  - Non-blacklisted files are _still cleaned_
  - Unsupported files are still triggering an error

refactor

atenart authored 6 years ago and jvoisin committed 6 years ago

Fix on small typo. Cosmetic patch.

Signed-off-by: Antoine Tenart <antoine.tenart@ack.tf>

Fedora defines the 'text/xml' mime type for xml files. Adds this mime
type to the harmless parser.

Fixes #36.

Signed-off-by: Antoine Tenart <antoine.tenart@ack.tf>