I was seeing some failures in travis ci where the build crashed because
it wasn't able to open a socket. This failure was happening in the
scalatest fork runner so I decided to avoid forking entirely. An
alternative approach would possibly be just to remove the scalatest
framework from the server test project but not forking is nicer anyway.
The old sbt launcher uses jansi 1.11, which is incompatible with jline3.
To work around this, we can use the jna terminal implementation for the
jline system terminal. This commit also switches to using the jline
TerminalBuilder for all system terminals except for the windows system
terminal with the thin client. The jline terminal builder uses
reflection that is difficult to make work with the thin client and it is
much easier to just manually construct the thin client. This is only
necessary for windows because on posix the thin client will fall back to
an implementation that shells out for stty commands.
It turns out that task progress actually introduces a fair bit of
overhead. The biggest issue is that the task progress callbacks block
the Execute main thread. This means that time in those callbacks
delays task evaluation, slowing down sbt. This was not negligible, I was
seeing a lot of the total time of a no-op compile in
https://github.com/jtjeferreira/sbt-multi-module-sample was spent in
TaskProgress callbacks. Prior to these changes, I ran 30 no-op compiles
in that project and the average time was about 570ms. This number got
worse and worse because there were memory leaks in the TaskProgress
object. After these changes, it dropped to 250ms and after jit-ing, it
would drop to about 200ms. I also successfully ran 5000 consecutive
no-op compiles without leaking any memory.
A lot of the overhead of task progress was in adding tasks to the
timings map in AbstractTaskProgress. Tasks were never removed and
ConcurrentHashMap insertion time is proportional to the size of the map
(not sure if it's linear, quadratic or other) which was why sbt actually
got slower and slower the longer it ran. Much of the time was spent
adding tasks to the progress timings.
To fix this, I did something similar to what I did to manage logger
state in https://github.com/jtjeferreira/sbt-multi-module-sample. In
MainLoop, we create a new TaskProgress instance before command
evaluation and clean it up after. Earlier I made TaskProgress an object
to try to ensure there was only one progress thread at a time, and that
introduced the memory leak. In addition to removing the leak, I was able
to improve performance by removing tasks from the timings map when they
completed. Unlike TaskTimings and TaskTraceEvent, we don't care about
tasks that have completed for TaskProgress so it is safe to remove them.
In addition to the memory leaks, I also reworked how the background
threads work. Instead of having one thread that sleeps and prints
progress reports, we now use two single threaded executors. One is a
scheduled executor that is used to schedule progress reports and the
other is the actual thread on which the report is generated. When
progress starts, we schedule a recurring report that is generated every
sleep interval until task evaluation completes. Whenever we add a new
task, if we have haven't previously generated a progress report, we
schedule a report in threshold milliseconds. If the task completes
before the threshold period has elapsed, we just cancel the schedule
report. By doing things this way, we reduce the total number of reports
that are generated. Because reports need to effectively lock System.out,
the less we generate them, the better.
I also modified the internal data structures of AbstractTaskProgress so
that there is a single task map of timings instead of one map for
timings and one for active tasks.
In order to make the console task work with scala 2.13 and the thin
client, we need to provide a way for the scala repl to use an sbt
provided jline3 terminal instead of the default terminal typically built
by the repl. We also need to put jline 3 higher up in the classloading
hierarchy to ensure that two versions of jline 3 are not loaded (which
makes it impossible to share the sbt terminal with the scala terminal).
One impact of this change is the decoupling of the version of
jline-terminal used by the in process scala console and the version
of jline-terminal specified by the scala version itself. It is possible
to override this by setting the `useScalaReplJLine` flag to true. When
that is set, the scala REPL will run in a fully isolated classloader. That
will ensure that the versions are consistent. It will, however, for sure
break the thin client and may interfere with the embedded shell ui.
As part of this work, I also discovered that jline 3 Terminal.getSize is
very slow. In jline 2, the terminal attributes were automatically cached with a
timeout of, I think, 1 second so it wasn't a big deal to call
Terminal.getAttributes. The getSize method in jline 3 is not cached and
it shells out to run a tty command. This caused a significant
performance regression in sbt because when progress is enabled, we call
Terminal.getSize whenever we log any messages. I added caching of
getSize at the TerminalImpl level to address this. The timeout is 1
second, which seems responsive enough for most use cases. We could also
move the calculation onto a background thread and have it periodically
updated, but that seems like overkill.
Prior to these changes, sbt was leaking large amounts of memory via
log4j appenders. sbt has an unusual use case for log4j because it
creates many ephemeral loggers while also having a global logger that is
supposed to work for the duration of the sbt session. There is a lot of
shared global state in log4j and properly cleaning up the ephemeral task
appenders would break global logging. This commit fixes the behavior by
introducing an alternate logging implementation. Users can still use the
old log4j logging implementation but it will be off by default. The
internal implementation is very simple: it just blocks the current
thread and writes to all of the appenders. Nevertheless, I found the
performance to be roughly identical to that of log4j in my sample
project. As an experiment, I did the appending on a thread pool and got
a significant performance improvement but I'll defer that to a later PR
since parallel io is harder to reason about.
Background: I was testing sbt performance in
https://github.com/jtjeferreira/sbt-multi-module-sample and noticed that
performance rapidly degraded after I ran compile a few times. I took a
heap dump and it became obvious that sbt was leaking console appenders.
Further investigation revealed that all of the leaking appenders in the
project were coming from task streams. This made me think that the fix
would be to track what loggers were created during task evaluation and
clear them out when task evaluation completed. That almost worked except
that log4j has an internal append only data structure containing logger
names. Since we create unique logger names for each run, that internal
data structure grew without bound. It looked like this could be worked
around by creating a new log4j Configuration (where that data structure
was stored) but while creating new configurations with each task runs
did fix the leak, it also broke global logging, which was using a
different configuration. At this point, I decided to write an alternate
implementation of the appender api where I could be sure that the
appenders were cleaned up without breaking global logging.
Implementation: I made ConsoleAppender a trait and made it no longer
extends log4j AbstractAppender. To do this, I had to remove the one
log4j specific method, append(LogEvent). ConsoleAppender now has a
method toLog4J that, in most cases, will return a log4j Appender that is
almost identical to the Appenders that we previously used. To manage
the loggers created during task evaluation, I introduce a new class,
LoggerContext. The LoggerContext determines which logging backend to use
and keeps track of what appenders and loggers have been created. We can
create a fresh LoggerContext before each task evaluation and clear it
out, cleaning up all of its resources after task evaluation concludes.
In order to make this work, there were many places where we need to
either pass in a LoggerContext or create a new one. The main magic is
happening in the `next(State)` method in Main. This is where we create a
new LoggerContext prior to command evaluation and clean it up after the
evaluation completes.
Users can toggle log4j using the new useLog4J key. They also can set the
system property, sbt.log.uselog4j. The global logger will use the sbt
internal implementation unless the system property is set.
There are a fairly significant number of mima issues since I changed the
type of ConsoleAppender. All of the mima changes were in the
sbt.internal package so I think this should be ok.
Effects: the memory leaks are gone. I successfully ran 5000 no-op
compiles in the sbt-multi-module-sample above with no degradation of
performace. There was a noticeable degradation after 30 no-op compiles
before.
During the refactor, I had to work on TestLogger and in doing so I also
fixed https://github.com/sbt/sbt/issues/4480.
This also should fix https://github.com/sbt/sbt/issues/4773
Using the scala reflect library always introduces significant
classloading overhead. We can eliminate the classloading overhead by
generating StringTypeTags at compile time instead.
This sped up average project loading time by a few hundred milliseconds
on my computer. The ManagedLoggedReporter in zinc is still using the
type tag based apis but after the next sbt release, we can upgrade the
zinc apis. We also could consider breaking binary compatibility.
sbt depends on scalacache (which hasn't been updated in about a year)
and we really don't need the functionality provided by scalacache. In
fact, the java api is somewhat easier to work with for our use case. The
motivation is that scalacache uses slf4j for logging which meant that it
was implicitly loading log4j. This caused some noisy logs during
shutdown when the previously unused cache was initialized just to be
cleaned up.
This commit also upgrades caffeine and moving forward we can always
upgrade caffeine (and potentially shade it) without any conflict with
the scalacache version.
This allows a user to install the native thin client into a particular
directory (e.g. /usr/local/bin). I also made buildNativeThinClient have
a file dependency on the classpath so that it can be incremental if the
classpath hasn't changed. This is useful if the user has run
buildNativeThinClient for testing and then decides to install it once
it's been validated without having to rebuild (which takes a minimum of
about 30 seconds on my laptop).
Ref https://github.com/sbt/zinc/pull/744
This implements `ThisBuild / usePipelining`, which configures subproject pipelining available from Zinc 1.4.0.
The basic idea is to start subproject compilation as soon as pickle JARs (early output) becomes available. This is in part enabled by Scala compiler's new flags `-Ypickle-java` and `-Ypickle-write`.
The other part of magic is the use of `Def.promise`:
```
earlyOutputPing := Def.promise[Boolean],
```
This notifies `compileEarly` task, which to the rest of the tasks would look like a normal task but in fact it is promise-blocked. In other words, without calling full `compile` task together, `compileEarly` will never return, forever waiting for the `earlyOutputPing`.
There were a number of unused key lint warnings when loading the sbt
build. In the case of `fork in compile` and `crossVersion in update`, it
wasn't clear that these were actually used, so I removed those settings.
The others seemed to be used so I just added them to the exclude list.
To make this work with legacy versions of sbt, I redefined the
excludeLintKeys key. Once we update the build.properties to a 1.4.x
version, we can drop the `val excludeLint` definition and replace
`excludeLint` with `excludeLintKeys`.
Side note: ++= does not work with excludeLintKeys which is why I used +=
for the excludes.
This commit upgrades sbt to using jline3. The advantage to jline3 is
that it has a significantly better tab completion engine that is more
similar to what you get from zsh or fish.
The diff is bigger than I'd hoped because there are a number of
behaviors that are different in jline3 vs jline2 in how the library
consumes input streams and implements various features. I also was
unable to remove jline2 because we need it for older versions of the
scala console to work correctly with the thin client. As a result, the
changes are largely additive.
A good amount of this commit was in adding more protocol so that the
remote client can forward its jline3 terminal information to the server.
There were a number of minor changes that I made that either fixed
outstanding ui bugs from #5620 or regressions due to differences between
jline3 and jline2.
The number one thing that caused problems is that the jline3 LineReader
insists on using a NonBlockingInputStream. The implementation ofo
NonBlockingInputStream seems buggy. Moreover, sbt internally uses a
non blocking input stream for system in so jline is adding non blocking
to an already non blocking stream, which is frustrating.
A long term solution might be to consider insourcing LineReader.java
from jline3 and just adapting it to use an sbt terminal rather than
fighting with the jline3 api. This would also have the advantage of not
conflicting with other versions of jline3. Even if we don't, we may want to
shade jline3 if that is possible.
The graalvm was swallowing all -D arguments and adding them to the
process system properties. This is undesirable since there are sbt
commands that have arguments starting with '-D'. It also breaks our
ability to pass system properties to the forked sbt process.
The existing implementation of watch did not work with the thin client.
In sbt 1.3.0, watch was changed to be a blocking command that performed
manual task evaluation. This commit makes the implementation more
similar to < 1.3.0 where watch modifies the state and after running the
user specified command(s), it enters a blocking command. The new
blocking command is very similar to the shell command.
As part of this change, I also reworked some of the internals of watch
so that a number of threads are spawned for reading file and input
events. By using background threads that write to a single event queue,
we are able to block on the file events and terminal input stream rather
than polling. After this change, the cpu utilization as measured by ps
drops from roughly 2% of a cpu to 0.
To integrate with the network client, we introduce a new UITask that is
similar to the AskUserTask but instead of reading lines and adding execs
to the command queue, it reads characters and converts them into watch
commands that we also append to the command queue.
With this new implementation, the watch task that was added in 1.3.0 no
longer works. My guess is that no one was really using it. It wasn't
documented anywhere. The motivation for the task implementation was that
it could be called within another task which would let users define a
task that monitors for file changes before running. Since this had never
been advertised and is only of limited utility anyway, I think it's fine
to break it.
I also had to disable the input-parser and symlinks tests. I'm not 100%
sure why the symlinks test was failing. It would tend to work on my
machine but fail in CI. I gave up on debugging it. The input-parser test
also fails but would be a good candidate to be moved to the client test
in the serverTestProj. At any rate, it was testing a code path that was
only exercised if the user changed the watchInputStream method which is
highly unlikely to have been done in any user builds.
The WatchSpec had become a nuisance and wasn't really preventing from
any regressions so I removed it. The scripted tests are how we test
watch.
This project is used to create client executables. The implementation is
pure java but we can build graalvm native-images from the java main
class. There are two versions of the client. One of them uses the
ipcsocket jni implementation to connect to the sbt server while the
other uses jna. It is necessary to use jni for the graalvm native-image
tool to work. Otherwise the two approaches should be identical.
This commit makes it possible for the sbt server to render the same ui
to multiple clients. The network client ui should look nearly identical
to the console ui except for the log messages about the experimental
client.
The way that it works is that it associates a ui thread with each
terminal. Whenever a command starts or completes, callbacks are invoked
on the various channels to update their ui state. For example, if there
are two clients and one of them runs compile, then the prompt is changed
from AskUser to Running for the terminal that initiated the command
while the other client remains in the AskUser state. Whenever the client
changes uses ui states, the existing thread is terminated if it is
running and a new thread is begun.
The UITask formalizes this process. It is based on the AskUser class
from older versions of sbt. In fact, there is an AskUserTask which is
very similar. It uses jline to read input from the terminal (which could
be a network terminal). When it gets a line, it submits it to the
CommandExchange and exits. Once the next command is run (which may or
may not be the command it submitted), the ui state will be reset.
The debug, info, warn and error commands should work with the multi
client ui. When run, they set the log level globally, not just for the
client that set the level.
Fixes https://github.com/sbt/sbt/issues/3112
This unpacks Extracted as State's extension methods.
In addition this provides a way of responding via LSP.
The repo overrides scripted test relies on using the launcher to modify
the default resolvers. To support this, I extended the scripted launcher
to use the bundled sbt launcher if it is passed in via the
`-Dsbt.launch.jar` system property.
xuwei-k
Ethan Atkins
Eugene Yokota
Adrien Piquerez
Brice Jaglin
Dale Wijnand