[metrics-bugs] #24042 [Metrics/Library]: Test descriptor parsing with fuzzed descriptors

[Tor Bug Tracker & Wiki]

#24042: Test descriptor parsing with fuzzed descriptors
---------------------------------+--------------------------
     Reporter:  karsten          |      Owner:  metrics-team
         Type:  enhancement      |     Status:  new
     Priority:  Medium           |  Milestone:
    Component:  Metrics/Library  |    Version:
     Severity:  Normal           |   Keywords:
Actual Points:                   |  Parent ID:
       Points:                   |   Reviewer:
      Sponsor:                   |
---------------------------------+--------------------------
 We recently learned the hard way that there exist descriptors that are
 valid according to the specification that our descriptor parsers have
 trouble with. We should throw a bunch of fuzzed descriptors into a local
 Metrics instance to see if anything breaks.

 This could be a trivial reader application using metrics-lib, or it could
 be a local Onionoo or metrics-web or ExoneraTor instance.

 I'm putting this into Metrics/Library for now, because that's likely be
 responsible for the bulk of descriptor-parsing issues, even though there's
 also some parsing code in other components.

 According to teor, Tor has a [https://gitweb.torproject.org/fuzzing-
 corpora.git/tree/ collection of valid and invalid directory documents].
 Part of the challenge here will be to tell valid and invalid documents
 apart. That might be a manual process.

 And according to atagar, Stem now supports
 [https://stem.torproject.org/tutorials/mirror_mirror_on_the_wall.html#can-i
 -create-descriptors descriptor creation] (and should have all capabilities
 Leekspin did). This is just in case we run out of descriptors above and
 feel like creating even more would be a good idea.

--
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/24042>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online