1. 13 Apr, 2019 1 commit
  2. 12 Apr, 2019 3 commits
  3. 11 Apr, 2019 5 commits
  4. 10 Apr, 2019 1 commit
  5. 08 Apr, 2019 1 commit
  6. 07 Apr, 2019 1 commit
    • yacovm's avatar
      [FAB-14944] Fix Data race in TestSend · 83799838
      yacovm authored
      
      
      The test has a data race because it re-programs a mock in
      a non memory coherent (and wishful thinking) way, and the
      goroutine that sends the messages isn't effected by
      the mock synchronization attempts in the test, because
      they happen after the data race occurs.
      
      This change set fixes the data race by programming the stream's
      mock only once, and having it dynamically read the returned values
      while holding a lock, which synchronizes between the goroutine
      that sends down the stream and the test goroutine.
      
      Ran this 10,000 times with data race detector and it doesn't
      fail now.
      
      Change-Id: I8a0030586da9102e8539c9f7b8699c63c6565894
      Signed-off-by: default avataryacovm <yacovm@il.ibm.com>
      83799838
  7. 06 Apr, 2019 1 commit
    • Yacov Manevich's avatar
      Merge changes from topics 'FAB-14239', 'FAB-13929' into release-1.4 · db088625
      Yacov Manevich authored
      * changes:
        [FAB-14328] de-vipererize gossip discovery
        [FAB-14322] de-vipererize gossip state
        [FAB-14295] de-vipererize gossip privdata
        [FAB-13534] cache pushAckTimeout in pvt data
        [FAB-14291] de-vipererize gossip election
        [FAB-14265] de-vipererize gossip comm
        [FAB-14264] de-vipererize gossip pull engine
        [FAB-14239] get rid of  NewCommInstanceWithServer
        [FAB-14238] use ephemeral ports in gossip state tests
        [FAB-14218] use ephemeral ports in gossip service tests
        [FAB-14184] use ephemeral ports in gossip gossip tests
      db088625
  8. 04 Apr, 2019 3 commits
  9. 03 Apr, 2019 15 commits
  10. 02 Apr, 2019 7 commits
  11. 01 Apr, 2019 2 commits
    • yacovm's avatar
      [FAB-13653] Fix TestTicker flake · fc10fabf
      yacovm authored
      
      
      TestTicker/Stop ticker serially (very) rarely fails
      due to scheduling (the goroutine starts emitting a tick
      right before the channel is closed, and therefore
      the test reads a value from the channel despite it
      being closed).
      
      Fixed it by using gomega.Eventually.
      
      Change-Id: I2088814d707939fa8d4d63116657178d30875776
      Signed-off-by: default avataryacovm <yacovm@il.ibm.com>
      (cherry picked from commit 333d1169)
      fc10fabf
    • yacovm's avatar
      [FAB-14818] Flake in TestStreamAbort · b05a1b7d
      yacovm authored
      
      
      This change set fixes a flake in TestStreamAbort by ensuring
      the server-side handler doesn't return before the stream
      is canceled by the reconfiguration.
      
      This is needed, because gRPC selects the error to return
      to the Recv() method by doing a select { } between
      the actual bytes read from the stream, and the context.
      
      In the test, we test that the context is canceled due to
      the reconfiguration, but we also concurrently let the server
      side handling of the request finish - it might be that the
      bytes read (and the corresponding EOF signal) happens before
      the context cancelling, and therefore the expected error
      is not received.
      
      By delaying the server side handling of the request, we ensure
      that we either abort the stream, or the test never ends.
      
      I ran it (with data race detector) 5,000 times and it succeeded.
      
      Change-Id: Ie1ecf360dde79c3afa7bf919a84a20b3b600c865
      Signed-off-by: default avataryacovm <yacovm@il.ibm.com>
      (cherry picked from commit 2cafd0af)
      b05a1b7d