This is a Q&D fix to get kubo building again, this is completely equivalent to what we used to have.
I'll submit a patch that DI the verifcid.Allowlist interface with fx later but this is needed to unblock something else.
Mplex does not implement backpressure, our implementation will randomly reset streams if buffers overflow instead of risking deadlocks.
In the past we had a bug where kubo nodes would prefer mplex over yamux. Turning off mplex make our connections to thoses nodes negociate yamux.
Closes#9958
* test(t0112): drop test ported to conformance
* test(t0113): drop test ported to conformance
* test(t0114): drop test ported to conformance
* test(t0114): drop test ported to conformance
* test(t0115): drop test ported to conformance
* test(t0122): drop test ported to conformance
* test(t0123): drop test ported to conformance
* test(t0117): drop test ported to conformance
* test(t0124): drop test ported to conformance
* test(t0114): simplify tests
* test(t0112): drop test ported to conformance
* test(t0116): drop test ported to conformance
* t0114: restore tests flagged by lidel
* t0112: restore
* t0116: restore dirindex check
* t0109: restore file
* t0115: restore full file
* t0114: restored rest of file
* fix: kill the iptb cluster
Streams used to be blocked on ping IO because we didn't handled the DHT ping check asynchronously.
Include fixes from libp2p/go-libp2p-kad-dht#851
Fixes#9957
* feat!: dag import - don't pin roots by default
Fixes: https://github.com/ipfs/kubo/issues/9765
* test(ipip-402): dag import
this adds basic regression test that guards behavior
around partial cars with or without pinning
* docs(ipip-402): ipip and dag import changelog
---------
Co-authored-by: Marcin Rataj <lidel@lidel.org>
* feat(bootstrap): save connected peers as backup temporary bootstrap ones
* fix: do not add duplicated oldSavedPeers, not using tags, reuse
randomizeList
* test: add regression test
* chore: add changelog
---------
Co-authored-by: Henrique Dias <hacdias@gmail.com>
Co-authored-by: Marcin Rataj <lidel@lidel.org>
GH Actions recently changed their Docker build implementation and it
has a different output than previously, causing the tests that parse
its output to fail.
This switches the test to not parse Docker build output. The parsing
was used to extract the image ID while still showing logs. A better
way to show logs and still know the image ID is to tag it, which is
what this now does.
This also renames the Docker tests so that they run earlier. This
takes better advantage of the fact that the sharness tests are run in
parallel. Since the Docker test are quite long, and are at the end of
the list, the test runner is not running other tests in parallel while
the Docker tests are running.
The multinode test is effectively the same as the twonode test. There
are some problems with it too: it *looks* like it's testing the
Websocket transport with the "listentype,ws" IPTB attribute, but that
attribute doesn't actually exist in ipfs/iptb-plugins, so it does
nothing, so that test actually just runs the same test twice (Yamux
disabled). Furthermore, this is just the same test as in the mplex
twonode test. So this just removes the useless multinode test
entirely.
Also, this removes the part of the twonode test that checks the amount
of data transferred over Bitswap. This is an implementation detail of
Bitswap, it's not appropriate to test this in an end-to-end test as it
depends on algorithmic details of how Bitswap works, and has nothing
to do with transports. This is probably more appropriate as a perf or
benchmark test of Bitswap.
This also moves equivalent functionality from jbenet/go-random-files
into the testutils package. This just copies the code and modifies it
slightly for better ergonomics.
