IF YOU WOULD LIKE TO GET AN ACCOUNT, please write an
email to Administrator. User accounts are meant only to access repo
and report issues and/or generate pull requests.
This is a purpose-specific Git hosting for
BaseALT
projects. Thank you for your understanding!
Только зарегистрированные пользователи имеют доступ к сервису!
Для получения аккаунта, обратитесь к администратору.
see comments in #458. it is unknown why this is needed but it does work
away of the compiler error.
it was later found out by bisecting Cargo.lock that this is caused by
async-trait 0.1.42 => 0.1.43 upgrade, but it's very unclear why this
would be caused by the only PR in that release.
468: chore(clippy) r=koivunej a=Mirko-von-Leipzig
Changes with clippy fixes. Split into smaller commits -- can be squashed if desired.
Co-authored-by: Mirko von Leipzig <mirko.vonleipzig@gmail.com>
469: Fix (windows) build r=koivunej a=koivunej
The old vcpkg commit `ffa41582f78478812c836a6e8ce315fb27431182` has started segfaulting for some unknown reason. This PR:
1. updates run-vcpkg, openssl-sys, openssl along with vcpkg revision `261c458af6e3eed5d099144aff95d2b5035f656b`
2. disables CI on mobile platforms for now
3. allows rustfmt and clippy to fail
(2) is to save CI seconds until we know better; when this was decided previously CI seconds were practically unlimited; we'll need another decision on that. I don't think anyone is using the android build for example yet.
(3) will follow up with another PR to handle these.
Co-authored-by: Joonas Koivunen <joonas.koivunen@gmail.com>
if you want to contest this change, please do go ahead and let us know
you'd benefit from having these enabled. originally they were added
because adding multiple images to build worked most of the time without
any downside with unlimited free CI.
457: Migrate dnslink to trust-dns r=koivunej a=koivunej
Ahead next version of libp2p-dns which will finally use trust-dns. Added FIXMEs and more importantly get rid of the windows specific ipconfig, custom future impl around selectok. This shouldn't really affect anything so I doubt this needs to be mentioned in changelog. The IPNS story is quite weak as it is.
Co-authored-by: Joonas Koivunen <joonas.koivunen@gmail.com>
456: Fix swarmapi tests and add more cases r=koivunej a=koivunej
This PR fixes the swarm tests and yet another swarmapi related panic reachable through only the swarmapi "unit tests". It also adds a two new cases which would had been quite ugly add to under `tests/`.
First of all, the test had been broken long time ago in a rust-libp2p update when the listening address adding became asynchronous.
The additional panic exposed by now working test case comes from the `inject_connected` never arriving when only using SwarmApi as the only behaviour. It uses the DummyProtocolHandler which opens no streams and demands keepalive, the connection is immediatedly closed.
The additional test cases cover connecting to wrong peerid, and dialing multiple addresses when only first one succeeds, the second errors.
Co-authored-by: Joonas Koivunen <joonas.koivunen@gmail.com>
Co-authored-by: Joonas Koivunen <joonas@equilibrium.co>
this allows not leaking the SubscriptionErr type anywhere more, but it
is sort of kicking the can on the way as we are starting to desperately
need proper error types instead of the stringified one.
the test case is rather special one: when the connection is established
there will be a disconnect right away. the unconditional subscription
finishing tripped up the "debug assert" in subscriptions. it might be a
bit wrong to end up with an `Ok(())` value with the test case, but that
is the outcome currently.
this was broken in some upgrade, not the most recent one. it cleverly
only looked like it was doing something but since starting to listen on
an address doesn't do anything but queue up the request, there is
nothing to do for the for-loop and so everything dies down. luckily we
might not need this async unit test here.
454: Fix connecting to peers r=koivunej a=koivunej
The correct way to connect to a peer is in my understanding to DialPeer and then proceed to give the address of that said peer from `NetworkBehaviour::addresses_of_peer`. While doing this, the previous workaround for missing events from banning a peer (the way we force a disconnect) was also necessary to fix.
Co-authored-by: Joonas Koivunen <joonas.koivunen@gmail.com>
455: test: split off common_tests from common r=koivunej a=koivunej
Tiny refactoring to stop testing `common::tests` as part of every top level `test/*.rs`, noticed during #454.
Co-authored-by: Joonas Koivunen <joonas.koivunen@gmail.com>
while libp2p might soon gain the ability to dial peer addresses
directly, this will be a stopgap until that. updated a bunch of comments
and adapted to the since fixed missing events on banning the peer, which
is still used to disconnect peer at will.
453: Fix multinode test spans r=koivunej a=koivunej
This PR changes how the spans are set up from IpfsOptions and in UninitializedIpfs. While debugging the now ignored test (see #450) I found that the spans were configured wrong and thus none of the multiple nodes created by `spawn_nodes` could be differentiated.
This also renames the spans to more logical from the "$root" given at `IpfsOptions::span`:
`Span::current` => `$root:init`
`Span::current` => `$root:init:swarm`
`$root:swarm` => `$root:exec` (literally the executor libp2p spawns futures through)
`$root` => `$root:bg_task` => `$root:swarm` (background task)
`$root` => `$root:facade` (futures created through Ipfs::* methods)
Still very far from perfect, but perhaps a step into better direction.
Co-authored-by: Joonas Koivunen <joonas.koivunen@gmail.com>
