1. add eof channel to event emitter and use this channel as signal
that we shouldn't be sending anymore instead of using a pattern where
we use a timer to decide sending has timed out (because we're using
a buffered channel, it is still possible for some evetns to end up
in the channel if there is space, but this is not a concern, because
the events will be deleted when the channel itself is gone);
2. refactor all tests where we assumed the output channel was closed
to actually use a parallel "eof" channel and use it as signal we
should not be sending anymore (not strictly required but still the
right thing to do in terms of using consistent patterns);
3. modify how we construct a runner so that it passes to the
event emitter an "eof" channel and closes this channel when the
main goroutine running the task is terminating;
4. modify the task to signal events such as "task goroutine started"
and "task goroutine stopped" using channels, which helps to write
much more correct tests;
5. take advantage of the previous change to improve the test that
ensures we're not blocking for a small number of events and also
improve the name of such a test to reflect what it's testing.
The related issue in term of fixing the channel usage pattern is
https://github.com/ooni/probe/issues/1438.
Regarding improving testability, instead, the correct reference
issue is https://github.com/ooni/probe/issues/1903.
There are possibly more changes to apply here to improve this package
and its testability, but let's land this diff first and then see
how much time is left for further improvements.
I've run unit and integration tests with `-race` locally.
This diff will need to be backported to `release/3.11`.
This diff forward ports 90bf0629b957c912a5a6f3bb6c98ad3abb5a2ff6 to `master`.
If we close the channel to signal the end of a task we may panic
when some background goroutine tries to post on the channel.
This bug is rare but may happen.
See for example https://github.com/ooni/probe/issues/1438.
How can we improve?
First, let us add a timeout when sending to the channel. Given that
the channel is buffered and we have a generous timeout (1/4 of a
second), it's unlikely we will really block. But, in the event in
which a late message appears, we'll eventually _unblock_ when
sending with a timeout. So, now we don't have to worry anymore
about leaking forever a goroutine.
Then, let us change the protocol with which we signal that a task
is done. We used to close the channel. Now, instead we just
synchronously post a nil on the channel when done.
In turn, we interpret this nil to mean that the task is done when
we receive messages.
The _main_ different with respect to before is that now we are
asking the consumer of our API to drain the channel. Because
before we had a blocking channel, it seems to me we were already
requiring the consumer of the API to do that. Which means, I think
in practical terms it did not change much.
Finally, acknowledge that we don't need a specific state variable
to tell us we're done and simplify a little bit the API by
just making isRunning private and using the "we're done" signal
to determine whether we've stopped running the task.
All these changes should be enough to close https://github.com/ooni/probe/issues/1438.
* refactor(atomicx): move outside the engine package
After merging probe-engine into probe-cli, my impression is that we have
too much unnecessary nesting of packages in this repository.
The idea of this commit and of a bunch of following commits will instead
be to reduce the nesting and simplify the structure.
While there, improve the documentation.
* fix: always use the atomicx package
For consistency, never use sync/atomic and always use ./internal/atomicx
so we can just grep and make sure we're not risking to crash if we make
a subtle mistake on a 32 bit platform.
While there, mention in the contributing guidelines that we want to
always prefer the ./internal/atomicx package over sync/atomic.
* fix(atomicx): remove unnecessary constructor
We don't need a constructor here. The default constructed `&Int64{}`
instance is already usable and the constructor does not add anything to
what we are doing, rather it just creates extra confusion.
* cleanup(atomicx): we are not using Float64
Because atomicx.Float64 is unused, we can safely zap it.
* cleanup(atomicx): simplify impl and improve tests
We can simplify the implementation by using defer and by letting
the Load() method call Add(0).
We can improve tests by making many goroutines updated the
atomic int64 value concurrently.
* refactor(fsx): can live in the ./internal pkg
Let us reduce the amount of nesting. While there, ensure that the
package only exports the bare minimum, and improve the documentation
of the tests, to ease reading the code.
* refactor: move runtimex to ./internal
* refactor: move shellx into the ./internal package
While there, remove unnecessary dependency between packages.
While there, specify in the contributing guidelines that
one should use x/sys/execabs instead of os/exec.
* refactor: move ooapi into the ./internal pkg
* refactor(humanize): move to ./internal and better docs
* refactor: move platform to ./internal
* refactor(randx): move to ./internal
* refactor(multierror): move into the ./internal pkg
* refactor(kvstore): all kvstores in ./internal
Rather than having part of the kvstore inside ./internal/engine/kvstore
and part in ./internal/engine/kvstore.go, let us put every piece of code
that is kvstore related into the ./internal/kvstore package.
* fix(kvstore): always return ErrNoSuchKey on Get() error
It should help to use the kvstore everywhere removing all the
copies that are lingering around the tree.
* sessionresolver: make KVStore mandatory
Simplifies implementation. While there, use the ./internal/kvstore
package rather than having our private implementation.
* fix(ooapi): use the ./internal/kvstore package
* fix(platform): better documentation
* refactor: start building an Android package
Part of https://github.com/ooni/probe/issues/1335.
This seems also a good moment to move some packages out of the
engine, e.g., oonimkall. This package, for example, is a consumer
of the engine, so it makes sense it's not _inside_ it.
* fix: committed some stuff I didn't need to commit
* fix: oonimkall needs to be public to build
The side effect is that we will probably need to bump the major
version number every time we change one of these APIs.
(We can also of course choose to violate the basic guidelines of Go
software, but I believe this is bad form.)
I have no problem in bumping the major quite frequently and in
any case this monorepo solution is convinving me more than continuing
to keep a split between engine and cli. The need to embed assets to
make the probe more reliable trumps the negative effects of having to
~frequently bump major because we expose a public API.
* fix: let's not forget about libooniffi
Honestly, I don't know what to do with this library. I added it
to provide a drop in replacement for MK but I have no idea whether
it's used and useful. I would not feel comfortable exposing it,
unlike oonimkall, since we're not using it.
It may be that the right thing to do here is just to delete the
package and reduce the amount of code we're maintaining?
* woops, we're still missing the publish android script
* fix(publish-android.bash): add proper API key
* ouch fix another place where the name changed
