Log for src/testing.rs

• Only compile TrivialID for codecs 💬 by Caio 2 years ago

  Only used for roundtrip tests
  

• promote AccumulatingRuntime to supported 💬 by Caio 2 years ago

  It's useful enough, lets me get rid of yet another dummy runtime impl
  and allows examples to be smaller
  

• New Notification::Rename for addr conflict resolution by Caio 2 years ago
• conflict resolution within Members via Identity 💬 by Caio 2 years ago

  I really don't like the `win_addr_conflict` name
  
  ... not like `has_same_prefix` was any better
  

• wip conflict management sketch 💬 by Caio 2 years ago

  sweet. looks like simple conflict mgmt will be a go
  

• Stable addr for testing ID 💬 by Caio 2 years ago

  huh it wasn't a proper breakage
  

• use addr equality instead of has_same_prefix 💬 by Caio 2 years ago

  finally something broke proper. soon i'll know if this will actually
  go somewhere
  

• identity with associated addr by Caio 2 years ago
• identity.addr() needs PartialEq, prolly not Hash by Caio 2 years ago
• wip Identity::addr() 💬 by Caio 2 years ago

  not sure this will pan out yet
  

• Fail compilation on unreachable public exports 💬 by Caio 2 years ago

  This enables the unreachable_pub rustc lint so that we don't end
  up exposing unconstructable types again
  
  Ref: https://github.com/caio/foca/issues/24
  

• bugfix: Feed messages can come with trailing junk 💬 by Caio 3 years ago

  When crafting a Message::Feed payload, we were assuming that codecs
  wouldn't leave ANY invalid data in the buffer when an error
  occurred
  
  This is very likely to happen when the messages being
  sent would reach Config::max_packet_size (i.e.: large clusters,
  large identities) and the usd codec is not careful checking if
  the whole Member can fit the buffer before writing anything to
  it.
  
  Whilst annoying, the errors this led to were harmless noise
  unless there's a chance a codec would interpret the trailing
  bytes as part of the previous entry in the buffer- quite
  unlikely I think?
  
  Before the fix, the added test fails with:
  
  > ---- tests::feed_does_not_contain_trailing_jumk stdout ----
  > thread 'tests::feed_does_not_contain_trailing_jumk' panicked at 'assertion failed: `(left == right)`
  >   left: `Ok(())`,
  >  right: `Err(CustomBroadcast(Broadcasts disabled))`', src/lib.rs:3634:9
  

• Ability to notify active down members 💬 by Caio 3 years ago

  When a member is declared down by the cluster, its messages get ignored
  until the cluster "forgets" it went down (i.e.: after
  `Config::remove_down_after` expires). This patch introduces a new
  setting `Config::notify_down_members` which makes foca send
  a lightweight message to said members when they're down.
  

• clippy: derive Eq when deriving PartialEq by Caio 3 years ago
• Functionality to guarantee broadcast dissemination 💬 by Caio 4 years ago

  This patch introduces `Foca::broadcast` which works similarly to
  `Foca::gossip`; It:
  
  1. Picks `Config::num_indirect_probes` members
  2. Sends as much broadcast data as it can for them
  
  The difference from `gossip` is that this guarantees broadcast
  dissemination if there's any active member that passes the
  `BroadcastHandler::should_add_broadcast_data` condition; And
  the message is `Message::Broadcast` (instead of `Message::Gossip`)
  to signal that there's only broadcast data in the payload, no
  cluster updates.
  

• Reproduce incorrect probe cycle cleanup scenario by Caio 4 years ago
• Initial public commit 💬 by Caio 4 years ago

  Probably good enough for a first experimental release