[tor-bugs] #26675 [Core Tor/Torflow]: Incorrect ancillory data presented on Torflow guard relay "update" vote lines (was: possible server bug in Torflow guard "update" vote calculation)

[Tor Bug Tracker & Wiki]

#26675: Incorrect ancillory data presented on Torflow guard relay "update" vote
lines
------------------------------+---------------------
 Reporter:  starlight         |          Owner:  tom
     Type:  defect            |         Status:  new
 Priority:  Medium            |      Milestone:
Component:  Core Tor/Torflow  |        Version:
 Severity:  Minor             |     Resolution:
 Keywords:                    |  Actual Points:
Parent ID:                    |         Points:
 Reviewer:                    |        Sponsor:
------------------------------+---------------------
Changes (by starlight):

 * severity:  Normal => Minor


Comment:

 So this is not a serious problem.

 I overlooked that copy_vote does not overwrite pid_error and that lines
 702-704

 {{{
 # Reset the remaining vote data..
 n.measured_at = prev_votes.vote_map[n.idhex].measured_at
 n.pid_error = prev_votes.vote_map[n.idhex].pid_error
 }}}

 do overwrite pid_error and measured_at after lines 684-688 calculate a
 complete new vote from current scan information.  Had reviewed with an
 assumption, based on observed vote records, that votes in this situation
 are revisions to earlier votes rather than fresh ones.

 When use_pid_tgt is 1 vote revision appears to be what occurs, but current
 operation with use_pid_tgt is 0 produces full vote calculations while
 explicitly removing all calculation fields aside from new_bw, reverting to
 earlier values up to two weeks old.

 The behavior has no negative effect on the critical bw= (new_bw) value,
 but the other fields show incorrect data.  Behavior is a defect though of
 minor consequence.

 This issue could be closed WONTFIX in light of plans to retire Torflow,
 but it might make sense to repair the glitch and deploy to one BWauth for
 improved usefulness of vote files in support of validation and comparison
 work for the new system.

 Likely can be repaired by adding an exclusion to the "is this a guard"
 test when use_pid_tgt is not in effect.

--
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/26675#comment:3>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online