- Adds a CheckStyle configuration which is pretty similar to CC's
existing one.
- Add the Gradle license plugin.
- Ensure the existing source code is compatible with these additional
checks.
See #239
- os.time, when given a table, will act the same as PUC Lua - returning
the seconds since the epoch. We preserve the previous string/nil
behaviour though - os.epoch("local") is equivalent to PUC's
os.time().
- os.date will now act accept a string and (optional) time, returning
an appropriate table.
Somewhat resolves the madness which was dan200/ComputerCraft#183, and
hopefully (though probably not) makes @Vexatos happy.
This changes the previous behaviour a little, but hopefully is more
sane:
- Only require the socket to be open when first calling receive. This
means if it closes while receving, you won't get an error.
This behaviour is still not perfect - the socket could have closed,
but the event not reached the user yet, but it's better.
- Listen to websocket_close events while receiving, and return null
should it match ours.
See #201
We were using += instead of =, meaning the budget always grew,
rather than growing while there was still space. As a result, computers
were never correctly rate limited.
Further more, if a computer went into a deficit, we would continue to
increase the budget by a negative amount, exponentially decreasing until
overflowing!
Yes, this is a very embarrassing mistake. I'd been aware that rate
limiting wasn't working as expected for a while, I hadn't realised
the problem would be this stupid.
I promise! The joys of using -SNAPSHOT I guess...
This will now correctly cause orphaned threads to be cleaned up,
reducing the risk of thread saturation.
Previously we just relied on magic int values, which was confusing and
potentially error-prone. We could use EnumFacing, but that's a)
dependent on MC and b) incorrect, as we're referring to local
coordinates.
Are most of these changes small and petty? Yes. However, IMO they do
make the code more readable. Anyway, a summary of some of the more
interesting changes:
- Expose Abstract*Upgrade classes in the API
- Fix the spelling of Jonathan in the API docs (*shakes fist*)
- Fix bug with printout not working in the offhand.
- Rename any argments/variables accidentally named "m_*", and add an
inspection to prevent it happening again.
- Remove most of the Block*.Properties classes - just inline them in
the parent class.
- Return super.writeToNBT instead of reassigning at the top.
OK, so let's get this out of the way, there's some actual changes mixed
in here too. I'm really sorry:
- Turtles can now not be renamed with unnamed item tags (previously it
would clear the name, this seemed a little unideal).
- commands.getBlock(s)Data will also include NBT.
Now, onto the horror story which is these inspection changes:
- Make a lot of methods static
- Typo fixes
- Make utility classes final + private constructor
- Lots of reformatting (ifs -> ternary, invert control flow, etc...)
- ???
- Profit!
I'm so going to regret this - can pretty much guarantee this is going to
break something.
This uses a similar approach to ComputerThread: executors store how long
they've spent executing tasks. We then use that time to prioritise
executors.
One should note that we use the current runtime at the point of adding
to the queue - external tasks will not contribute towards it until a
later execution.
This effectively acts as a public interface to canExecuteExternal() and
consumeTime(). It's hopefully sufficiently general that we can mess
around with the backend as much as we like in the future.
One thing to note here is that this is based on a polling API, as it's
largely intended for people running work every tick. It would be
possible to adapt this with callbacks for when work is available,
etc..., but that was not needed immediately.
This also removes IComputerOwned, as Plethora no longer needs it.
Unlike ComputerThread, we do not have a single source of tasks, and so
need a smarter way to handle scheduling and rate limiting. This
introduces a cooldown system, which works on both a global and
per-computer level:
Each computer is allowed to do some work for 5ms. If they go over that
budget, then they are marked as "hot", and will not execute work on the
next tick, until they have cooled down. This ensures that _on average_
computers perform at most 5ms of work per tick.
Obviously this is a rather large time span, so we also apply a global
10ms to all computers. This uses the same cooldown principle, meaning we
keep to an average of 10ms, even if we go over budget.
We attempted to simplify this 0bfb7049b0,
but that change now means that minimumVirtualRuntime is not updated. As
a result, new tasks will have a runtime of 0 when the queue is empty.
Oh goodness, this is going to painful to update to 1.13.
We now translate:
- Computer/Disk ID tooltips
- /computercraft descriptions, synopsises and usages. The last of these
may not always be translated when in SMP, as it is sometimes done on
the server, but the alternative would be more complex than I'm happy
with.
- Tracking field names. Might be worth adding descriptions too in the
future.
Also cleanup a couple of other translation keys, so they're more
consistent with Minecraft.
Closes#141
- Share the ILuaContext across all method calls, as well as shifting it
into an anonymous class.
- Move the load/loadstring prefixing into bios.lua
- Be less militant in prefixing chunk names:
- load will no longer do any auto-prefixing.
- loadstring will not prefix when there no chunk name is supplied.
Before we would do `"=" .. supplied_program`, which made no sense.
- Only update all runtimes and the minimum runtime when queuing new
exectors. We only need to update the current executor's runtime.
- Fix overflows when comparing times within TimeoutState.
System.nanotime() may (though probably won't) return negative values.
- Hopefully explain how the scheduler works a little bit.
- Runners would set their active executor before starting resetting the
time, meaning it would be judged as running and terminated.
- Similarly, the cumulative time start was reset to 0, meaning the
computer had been judged to run for an impossibly long time.
- If a computer hit the terminate threshold, but not the hard abort
one, then we'd print the stack trace of the terminated thread - we
now do it before interrupting.
There's still race conditions here when terminating a computer, but
hopefully these changes will mean they never occur under normal
operations (only when a computer has run for far too long).
- Fix the timeout error message displaying utter rot.
- Don't resize the runner array. We don't handle this correctly, so
we shouldn't handle it at all.
- Increment virtualRuntime after a task has executed.
- The computer queue is a priority queue sorted by "virtual runtime".
- Virtual runtime is based on the time this task has executed, divided
by the number of pending tasks.
- We try to execute every task within a given period. Each computer is
allocated a fair share of that period, depending how many tasks are
in the queue. Once a computer has used more than that period, the
computer is paused and the next one resumed.
TimeoutState now introduces a TIMESLICE, which is the maximum period of
time a computer can run before we will look into pausing it.
When we have executed a task for more than this period, and if there are
other computers waiting to execute work, then we will suspend the
machine.
Suspending the machine sets a flag on the ComputerExecutor, and pauses
the "cumulative" time - the time spent handling this particular event.
When resuming the machine, we restart our timer and resume the machine.
Oh goodness, when will it end?
- Computer errors are shown in red.
- Lua machine operations provide whether they succeeded, and an
optional error message (reason bios failed to load, timeout error,
another Lua error), which is then shown to the user.
- Clear the Cobalt "thrown soft abort" flag when resuming, rather than
every n instructions.
- Computers will clear their "should start" flag once the time has
expired, irrespective of whether it turned on or not. Before
computers would immediately restart after shutting down if the flag
had been set much earlier.
Errors within the Lua machine are displayed in a more friendly
When closing a BufferedWriter, we close the underlying writer. As we're
using channels, this is an instance of sun.nio.cs.StreamEncoder. This
will attempt to flush the pending character.
However, if throwing an exception within .write errors, the flush will
fail and so the underlying stream is not closed. This was causing us to
leak file descriptors.
We fix this by introducing ChannelWrappers - this holds the wrapper
object (say, a BufferedWriter) and underlying channel. When closed, we
dispose of the wrapper, and then the channel. You could think of this as
doing a nested try-with-resources, rather than a single one.
Note, this is not related to JDK-6378948 - this occurs in the underlying
stream encoder instead.
- TimeoutState uses nanoseconds rather than milliseconds. While this is
slightly less efficient on Windows, it's a) not the bottleneck of Lua
execution and b) we need a monotonic counter, otherwise we could
fail to terminate computers if the time changes.
- Add an exception handler to all threads.
- Document several classes a little better - I'm not sure how useful
all of these are, but _hopefully_ it'll make the internals a little
more accessible.
- Move state management (turnOn, shutdown, etc...) event handling and
the command queue into a ComputerExecutor
- This means the computer thread now just handles running "work" on
computer executors, rather than managing a separate command queue +
requeuing it.
- Instead of setting soft/hard timeouts on the ILuaMachine, we instead
provide it with a TimeoutState instance. This holds the current abort
flags, which can then be polled within debug hooks.
This means the Lua machine has to do less state management, but also
allows a more flexible implementation of aborts.
- Soft aborts are now handled by the TimeoutState - we track when the
task was started, and now only need to check we're more than 7s since
then.
Note, these timers work with millisecond granularity, rather than
nano, as this invokes substantially less overhead.
- Instead of having n runners being observed with n managers, we now
have n runners and 1 manager (or Monitor).
The runners are now responsible for pulling work from the queue. When
the start to execute a task, they set the time execution commenced.
The monitor then just checks each runner every 0.1s and handles hard
aborts (or killing the thread if need be).
- Rename unload -> close to be a little more consistent
- Make pollAndResetChanged be atomic, so we don't need to aquire a lock
- Get the computer queue from the task owner, rather than a separate
argument.
Ideally we'd add a couple more tests in the future, but this'll do for
now.
The bootstrap class is largely yoinked from CCTweaks-Lua, so is a tad
ugly. It works though.
Prior to this change we would schedule a new task which attached
peripherals on the ComputerThread on the empty task queue. This had a
couple of issues:
- Slow running tasks on the computer thread could result in delays in
peripherals being attached (technically, though rarely seen in
practice).
- Now that the ComputerThread runs tasks at once, there was a race
condition in computers being turned on/off and peripherals being
attached/detached.
Note, while the documentation said that peripherals would only be
(at|de)tached on the computer thread, wired modems would attach on the
server thread, so this was not the case in practice.
One should be aware that peripherals are still detached on the
computer thread, most notably when turning a computer on/off.
This is almost definitely going to break some less well-behaved mods,
and possible some of the well behaved ones. I've tested this on SC, so
it definitely works fine with Computronics and Plethora.
The Computer class currently has several resposiblities such as storing
id/label, managing redstone/peirpherals, handling management of the
computer (on/off/events) and updating the output.
In order to simplify this a little bit, we move our IAPIEnvironment
implementation into a separate file, and store all "world state"
(redstone + peripherals) in there. While we still need to have some
level of updating them within the main Computer instance, it's
substantially simpler.
- Fire close events instead of failure when open websockets error.
- Handle ping events. I thought I was doing this already, but this
requires a WebsocketProtocolHandler. Fixes#118
The latest version of Cobalt has several major changes, which I'm
looking forward to taking advantage of in the coming months:
- The Lua interpreter has been split up from the actual LuaClosure
instance. It now runs multiple functions within one loop, handling
pushing/popping and resuming method calls correctly.
This means we have a theoretically infinite call depth, as we're no
longer bounded by Java's stack size. In reality, this is limited to
32767 (Short.MAX_VALUE), as that's a mostly equivalent to the limits
PUC Lua exposes.
- The stack is no longer unwound in the event of errors. This both
simplifies error handling (not that CC:T needs to care about that)
but also means one can call debug.traceback on a now-dead coroutine
(which is more useful for debugging than using xpcall).
- Most significantly, coroutines are no longer each run on a dedicated
thread. Instead, yielding or resuming throws an exception to unwind
the Java stack and switches to a different coroutine.
In order to preserve compatability with CC's assumption about LuaJ's
threading model (namely that yielding blocks the thread), we also
provide a yieldBlock method (which CC:T consumes). This suspends the
current thread and switches execution to a new thread (see
SquidDev/Cobalt@b5ddf164f1 for more
details). While this does mean we need to use more than 1 thread,
it's still /substantially/ less than would otherwise be needed.
We've been running these changes on SwitchCraft for a few days now and
haven't seen any issues. One nice thing to observe is that the number of
CC thread has gone down from ~1.9k to ~100 (of those, ~70 are dedicated
to running coroutines). Similarly, the server has gone from generating
~15k threads over its lifetime, to ~3k. While this is still a lot, it's
a substantial improvement.
- Convert terminals from a polling-based system to a more event-driven
one: they now accept an onChanged callback, which marks the parent as
dirty.
- Schedule ticks when monitors are marked as dirty.
- Add several missing @Overrides. This has nothing to do with the rest
of the changes, but I'm bad at good git practice.
FileSystemMount was originally added to allow using ReadableByteChannels
instead of InputStreams. However, as zip files do not allow seeking,
there is no benefit of using them over the original JarMount (which we
need to preserve for backwards compatibility).
Instead of maintaining two near-identical mounts, we remove the
FileSystemMount and rewrite the JarMount implementation with several
improvements:
- Rewrite the jar scanning algorithm to be closer to 1.13+'s data pack
mount. This means we no longer require the jar file to have
directories before the file (though this was not a problem in
practice).
- Add all JarMounts to a ReferenceQueue, closing up the ZipFile when
they have been garbage collected (fixes#100).
- Cache the contents of all files for 60 seconds (with some constraints
on size). This allows us to seek on ROM files too (assuming they are
small), by reading the whole thing into memory.
The cache is shared across all mounts, and has a 64MiB limit, and
thus should not have an adverse impact on memory.
