This probably fails "responsible disclosure", but it's not an RCE and
frankly the whole bug is utterly hilarious so here we are...
It's possible to open a file on a disk drive and continue to read/write
to them after the disk has been removed:
local disk = peripheral.find("drive")
local input = fs.open(fs.combine(disk.getMountPath(), "stream"), "rb")
local output = fs.open(fs.combine(disk.getMountPath(), "stream"), "wb")
disk.ejectDisk()
-- input/output can still be interacted with.
This is pretty amusing, as now it allows us to move the disk somewhere
else and repeat - we've now got a private tunnel which two computers can
use to communicate.
Fixing this is intuitively quite simple - just close any open files
belonging to this mount. However, this is where things get messy thanks
to the wonderful joy of how CC's streams are handled.
As things stand, the filesystem effectively does the following flow::
- There is a function `open : String -> Channel' (file modes are
irrelevant here).
- Once a file is opened, we transform it into some <T extends
Closeable>. This is, for instance, a BufferedReader.
- We generate a "token" (i.e. FileSystemWrapper<T>), which we generate
a week reference to and map it to a tuple of our Channel and T. If
this token is ever garbage collected (someone forgot to call close()
on a file), then we close our T and Channel.
- This token and T are returned to the calling function, which then
constructs a Lua object.
The problem here is that if we close the underlying Channel+T before the
Lua object calls .close(), then it won't know the underlying channel is
closed, and you get some pretty ugly errors (e.g. "Stream Closed"). So
we've moved the "is open" state into the FileSystemWrapper<T>.
The whole system is incredibly complex at this point, and I'd really
like to clean it up. Ideally we could treat the HandleGeneric as the
token instead - this way we could potentially also clean up
FileSystemWrapperMount.
BBut something to play with in the future, and not when it's 10:30pm.
---
All this wall of text, and this isn't the only bug I've found with disks
today :/.
Name a more iconic duo than @SquidDev and over-engineered test
frameworks.
This uses Minecraft's test core[1] plus a home-grown framework to run
tests against computers in-world.
The general idea is:
- Build a structure in game.
- Save the structure to a file. This will be spawned in every time the
test is run.
- Write some code which asserts the structure behaves in a particular
way. This is done in Kotlin (shock, horror), as coroutines give us a
nice way to run asynchronous code while still running on the main
thread.
As with all my testing efforts, I still haven't actually written any
tests! It'd be good to go through some of the historic ones and write
some tests though. Turtle block placing and computer redstone
interactions are probably a good place to start.
[1]: https://www.youtube.com/watch?v=vXaWOJTCYNg
ForgeGradle (probably sensibly) yells at me about doing this. However:
- There's a reasonable number of mods doing this, which establishes
some optimistic precedent.
- The licence update in Aug 2020 now allows you to use them for
"development purposes". I guess source code counts??
- I'm fairly sure this is also compatible with the CCPL - there's an
exception for Minecraft code.
The main motivation for this is to make the Fabric port a little
easier. Hopefully folks (maybe me in the future, we'll see) will no
longer have to deal with mapping hell when merging - only mod loader
hell.
.getMethods() may throw if a method references classes which don't exist
(such as client-only classes on a server). This is an Error, and so is
unchecked - hence us not handling it before.
Fixes#645
- Fix doc library-path
- Only style <pre> code blocks as executable. Skip <code> ones.
- Document the default parameters in gps. Yes, we should do it
everywhere, but one has to start somewhere!
This has been broken for almost a year (28th Jan 2020), and I never
noticed. Good job me.
Fixes#641, closes#648 (basically the same, but targetting 1.15.x)
- Generate theme-color. Hopefully this time it works!
- Specify a site url. Technically this is wrong (we should use the
current git branch), but it's good enough for now.
- Move some options into a sub-category.
I knew I shouldn't do modding on things which aren't my main computer.
I actually did run checkstyleMain before committing, but entirely forgot
about this one. Go me.
- Move some common upgrade code to IUpgradeBase. 99% sure this this
preserves binary compatibility (on the JVM at least).
- Instead of requiring the share tag to match, allow upgrades to
specify their own predicate. IMO this is a little ugly, but required
to fix#614 as other mods chuck their own NBT on items.
- Remove incorrect impostor recipes for pocket computers. We were
generating them from the list of turtle upgrades instead!
- Fix JEI plugin not blocking impostor recipes as of the data-generator
rewrite.
As explained in the comment, "built-in" rendering types are now manually
rendered ("finish"ed) before calling finish() on the main renderer. This
means our depth blocker hasn't actually been drawn (if a monitor is the
last TE to be rendered), and so blocks pass the depth test when they
shouldn't.
Fixes#599
Maybe the capability system was a mistake in retrospect, as we don't
store the peripheral outside, so there's no way to reuse it. That will
probably come in a later change.
As a smaller fix, we pass the invalidate listener directly. The lifetime
of this is the same as the computer, so we don't create a new one each
time.
There's still the potential to leak memory if people break/replace a
computer (as listeners aren't removed), but that's an unavoidable flaw
with capabilities.
Fixes#593
