Dequeued fetches were always fetching all refs, instead of fetching the
specific refs that were initially requested. We fix that by storing the
refs to fetch in the fetch queue.
We were not properly guarding against the fact that we could receive
multiple sync events from the same node. Normally though, this wouldn't
happen, as further announcements would be stale.
Instead of using tuples and lots of arrays, this commit tries to use
`BTreeMap`s to provide more context to review, issues, patch and other
comments.
Also the returned object by the httpd is more aligned with what is
needed by the consumer, so less recomputation is needed for display.
When using `network: "test"`, we shouldn't default to any public seed
node. This change makes sure that the "preferredSeeds" configuration var
is only given a value when `network` is `main`.
The previous change to the wire protocol introduced a bug: outbound
peers that failed to connect fully would not notify the service on
disconnect. This change fixes it.
The command now works with both a replica target *and* a seeds target.
This is especially useful to eg. ensure that a preferred seed has been
synced, while still trying to hit a higher replica count.
* Make sure we cleanup pending peer states when disconnecting before
handshake.
* Make sure we don't panic if a peer disconnects before being fully
connected.
Introduce pinned repositories to the `profile::Config`, i.e.
`config.json`.
This is used for pinning repositories in the httpd `/projects` handle.
To allow returning all repositories, an `all` query parameter is also
included.
This is non-backwards compatible change, since we introduce the `show`
query parameter which, if not supplied, will default to `pinned`.
Signed-off-by: Fintan Halpenny <fintan.halpenny@gmail.com>
X-Clacks-Overhead: GNU Terry Pratchett
In the case where we don't find any seeds, display a better error
message.
We also don't bail out on error, since the announcement step might still
be worth trying.
When evaluating a COB, expose to the `apply` function the operations
concurrent to the current one being applied. These are essentially
"related" nodes in the graph, that are neither ancestors or descendants.
This will let us for example have different failure modes depending on
whether there were conflicting operations, as well as expose conflicts
to the user.
Previously, COB change commits would not distinguish between their
parents. This means that they would try to load source code commits
(eg. for patch COBs) and then fail since those commits wouldn't have a
manifest or `Rad-Resource`.
We introduce a backwards compatible change here which is that parent
oids that are not COB commits are accompanied by a commit trailer called
`Rad-Related`. This is for all related content that should be pulled in
but is not part of the change graph proper.
When building the change graph, we simply ignore commits that are marked
as "related".
We were using the two pretty interchangeably. "Project" should only be
used to refer to the repository payload. "Repository" should be used
when referring to the resource that is fetched, synced, cloned, checked
out etc.
When a DNS error occured, the `dial` call failed and we didn't get a
connection attempt, just a disconnect. We fix that by always calling
back into the service when we attempt a connection.
It is possible that there can be errors while configuring a Radicle
`Repository` during `rad::init`, for example:
- configuring working copy remote fails
- signing references fails
- setting the identity head fails
- setting the project head fails
If any of these occur, a best effort is made to remove the
`Repository` and the working copy remote, so that the caller may fix
any issues and retry.
Signed-off-by: Fintan Halpenny <fintan.halpenny@gmail.com>
X-Clacks-Overhead: GNU Terry Pratchett
Due to only using the `options.listen` address in the creation of the
`auth_url` we aren't able to authenticate with an existing httpd instance
if the instance isn't the default value.
Add a test to show the behaviour of setting up 3 delegates, through two
seed nodes -- ensuring that commands are still operational after
accepting a delegate that the node is missing.
The two seeds are setup with two different scopes:
- `Scope::All`
- `Scope::Followed`
This is to show that the network has eventual consistency once there
is a way to fetch any other node's fork in the network.
The setup of the test is:
- Alice adds Bob as a delegate
- Bob adds Eve as a delegate
- Alice accepts Bob's changes
Alice is able to get Eve's fork through the `Scope::All` seed. The
`Scope::Followed` seed will then be able get Eve's fork once any other
node syncs with it, in this case Eve.
Signed-off-by: Fintan Halpenny <fintan.halpenny@gmail.com>
X-Clacks-Overhead: GNU Terry Pratchett
The following scenario was noticed:
If a node has a `rad/id` that is behind the one it is fetching from,
where a delegate was added, then verification will fail since it will
not fetch the new delegate -- if the scope is set to `Followed`.
Irregardless if the `rad/id` is ahead or behind, the reference should
be fetched to use as the delegates anchor since it will tell the
fetching side which delegates in can fetch from the remote. If any of
those delegates are missing then the remote is indeed lying.
This is safe since the `refs/rad/id` reference is only updated by
computing the COB rather than setting the OID, so it will not update
through the fetch protocol itself.
Signed-off-by: Fintan Halpenny <fintan.halpenny@gmail.com>
X-Clacks-Overhead: GNU Terry Pratchett
Add a `Config` constructor for testing seed nodes that uses
`usize::MAX` for capacity.
This allows test scenarios to not drop messages due to rate limiting
-- all nodes are on the same IP address and rate limiting is performed
per IP address.
Signed-off-by: Fintan Halpenny <fintan.halpenny@gmail.com>
X-Clacks-Overhead: GNU Terry Pratchett
If the expected replica count had exceeded the amount of seeds already
synced with the following message would appear:
Not connected to any seeds for {rid}.
This is not true, since there are seeds that were already synced.
This is fixed by calculating the `synced` and `connected` nodes,
without the local peer in the set. This means that the check for
`synced.len() >= connected.len()` will get a hit and the function will
correctly return early.
Signed-off-by: Fintan Halpenny <fintan.halpenny@gmail.com>
X-Clacks-Overhead: GNU Terry Pratchett
We cannot guarantee that the `rad sync status` will run before all the
nodes are synced.
Remove the call to `rad sync status` and explain that at least 1 of
the nodes will be synced.
Signed-off-by: Fintan Halpenny <fintan.halpenny@gmail.com>
X-Clacks-Overhead: GNU Terry Pratchett
cloudhead
Sebastian Martinez
Fintan Halpenny
Lars Wirzenius