[tor-bugs] #19931 [Core Tor/Tor]: curve25519-donna runtime math errors on armhf

[Tor Bug Tracker & Wiki]

#19931: curve25519-donna runtime math errors on armhf
-------------------------------------------------+-------------------------
 Reporter:  chadmiller                           |          Owner:
     Type:  defect                               |         Status:  new
 Priority:  Medium                               |      Milestone:  Tor:
                                                 |  0.2.???
Component:  Core Tor/Tor                         |        Version:  Tor:
                                                 |  0.2.8.6
 Severity:  Normal                               |     Resolution:
 Keywords:  regression, integer-overflow,        |  Actual Points:
  029-proposed                                   |
Parent ID:                                       |         Points:  0.5
 Reviewer:                                       |        Sponsor:
-------------------------------------------------+-------------------------

Comment (by teor):

 This looks like a gcc bug where it doesn't implement -fwrapv correctly on
 your platform. (Or, at least, it implements it in a way that still results
 in warnings.)

 But it looks like curve25519 is still giving you the correct results.
 Please run the tor unit tests using 'make test' as an extra check that you
 are getting the correct results from curve25519.

 And here are your options:
 * ignore the errors
 * turn the error logging off for tor (or for your kernel)
 * find and use the compiler option that actually makes your compiler
 implement -fwrapv semantics without logging those warnings. If you find
 it, please tell us.
 * use another library that implements curve25519:
 {{{
 checking crypto_scalarmult_curve25519.h usability... no
 checking crypto_scalarmult_curve25519.h presence... no
 checking for crypto_scalarmult_curve25519.h... no
 checking nacl/crypto_scalarmult_curve25519.h usability... no
 checking nacl/crypto_scalarmult_curve25519.h presence... no
 checking for nacl/crypto_scalarmult_curve25519.h... no
 checking for nacl compiled with a fast curve25519 implementation... no
 }}}

 Also, as an aside, please run `ntpdate` every few hours to set your clock.
 (Or run the `ntp` daemon. It's more accurate, but it's also another
 network service, which has security and resource implications.)

    Aug 17 20:52:28 localhost Tor[2581]: Our clock is 7 minutes, 35 seconds
 behind the time published in the consensus network status document
 (2016-08-17 21:00:00 UTC). Tor needs an accurate clock to work correctly.
 Please check your time and date settings!

--
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/19931#comment:4>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online