GitLab

fix remaining issues for race detector in peer manager

Keep all of disconnection in a mutex

repace get/set with getOrCreate to keep operations atomic

Constrain use of mutex to actual operations on the peerQueues map via utility functions

Breakup Startup function so that wantlists are not sent with each call to SendMessage

If wantlist changes are present, send them immediately on startup, rather than as a seperate message

Limit retrying sending of a message even when a successful reconnect occurs

fix race conditions while setting up wantlists by creating peer queues on demand

BREAKING CHANGE: PeerManager SendMessage signature changed

fix #51

fix multiple data races

Ensures that we don't leave goroutines behind, even if the client forgets to
unsubscribe.

Calling `wg.Add` after `wg.Wait` has returned is invalid. This change swaps the
wait group for a plain rwmutex.

(caught with the race detector)

ci: add travis

Jenkins is EOL at this point.

providers: don't add every connected node as a provider

We now do exactly what the comment is warning about: track peers providing keys.

refactor(GetBlocks): Merge session/non-session

Make Bitswap GetBlocks just create a temporary session and use that code

fix #52 fix #49

Feat: A more robust provider finder for sessions (for now) and soon for all bitswap

Keep channels unblocked in cancelling request -- refactored to function. Also cancel find provider
context as soon as it can be.

Make sure if all requestors cancel their request to find providers on a peer, the overall query gets
cancelled

removed session id user completely from providerquerymanager

Removed a minor condition check that could fail in some cases just due to timing, but not a code
issue

Add debug logging for the provider query manager and make tests more reliable

Integrate the ProviderQueryManager into the SessionPeerManager and bitswap in general

re #52, re #49

Add functionality to timeout find provider requests so they don't run forever

Add a manger for querying providers on blocks, in charge of managing requests, deduping, and rate
limiting

fix(tests): stabilize session tests

Improve stability of tests for Session and SessionPeerManager

fix #61

contexts: make sure to abort when a context is canceled

Also, buffer single-use channels we may walk away from. This was showing
up (rarely) in a go-ipfs test.

fix(sessions): explicitly connect found peers

Remove sync.waitGroup in SessionPeerManager till it's needed