In the past, we were using `Focus.handleHelp()` to see if we're handling a
`help` command, and print the commands supported by the handler. We also used
`strcmp_P` directly to compare (parts of) our input against command supported by
the handler. This approach works, but had multiple major disadvantages: it
duplicated strings between `handleHelp` and the `strcmp_P` calls, and it relied
on fragile substring pointers to save space.
To replace all that, this patch implements a different approach. Help handling
is split between a check (`Focus.inputMatchesHelp()`) and a
reply (`Focus.printHelp()`), the latter of which takes a list of `PSTR()`
strings, rather than one single string. This allows us to reuse the same
strings for comparing against the handler's input.
The new approach no longer uses the fragile substring pointers, nor does it use
`strcmp_P` directly, but goes through a wrapper (`Focus.inputMatchesCommand()`)
instead.
These changes lead to a more readable pattern. While we do use longer strings as
a result, there is less duplication, and the new patterns also require less
code, so we end up with saving space, at least on AVR devices.
The old methods are still available and usable, but they're deprecated.
Signed-off-by: Gergely Nagy <algernon@keyboard.io>
`Serial.read()` is unbuffered, and will return `-1` (which we cast to an
unsigned char, so 255) immediately if there is no data in the incoming buffer.
This is unlike every other kind of read we do, which use `parseInt()`, and are
thus buffered reads with a timeout.
The problem with returning `-1` immediately is that Chrysalis sends data in
chunks, so if we end up trying to read a char at a chunk boundary, we'll end up
reading -1s, which we treat as valid data, and cast it to an unsigned char,
completely throwing off the protocol in the process.
By using `readBytes()`, we have a one second window during which more data can
arrive, and as such, is consistent with the rest of our reads.
Signed-off-by: Gergely Nagy <algernon@keyboard.io>
Previously, a newline that wasn't read with the rest of the command in
a single event would get appended to the command buffer instead of
remaining in the "peek" buffer, so the command dispatch would never
fire.
Stash a space delimiter instead of storing it in the buffer, only to
overwrite it later.
Fix an off-by-one error that could cause a buffered command to not be
null-terminated. This probably didn't cause a problem in practice,
because Focus plugins mostly did a `strcmp` against a fixed string that
is smaller than the command buffer size.
Signed-off-by: Taylor Yu <tlyu@mit.edu>
We should be treating the end of data the same way we treat a newline.
A common issue that affects a good number of plugins is that we don't deal with
trailing space well. For example, most plugins that respond with a large list of
values, where we iterate over an array or list, or something else, they usually
end up responding with a trailing space before the newline.
If we feed that same string back as an update, we can end up in a situation
where we lock up (or become very, very slow), because we want to read more data
than is available. Why? Because `Serial.parseInt()` (used by Focus under the
hood) will swallow up any leading whitespace. So if we have "255 \n" as an
argument list, we'll parse the first number, and the second `parseInt()` will
return 0, because it times out waiting for a number, consuming both the space
and the newline in the process. Thus, the next `::isEOL()` will still return
false, because `peek()` returns `-1`, signaling no data.
That can confuse the heck out of our plugins. To combat that, we should treat
end of data the same as EOL, and return false if `peek()` returns -1, too.
Signed-off-by: Gergely Nagy <algernon@keyboard.io>
This is the result of running the `include-what-you-use` wrapper, followed by
the `clang-format` wrapper on the Kaleidoscope codebase. It is now safe to use
both without needed any manual corrections after the fact, but it's still
necessary to run clang-format after IWYU, because the two differ in the way they
indent comments after header files.
Signed-off-by: Michael Richters <gedankenexperimenter@gmail.com>
Tested to work on both a Model 01 and a Model 100. This will become
unnecessary when the GD32 Arduino core grows the ability to autoflush on
SOF packets from the host
This standardizes namespace closing brackets for namespace blocks. Each one is
on its own line, with a comment clearly marking which namespace it closes.
Consecutive lines closing namespace blocks have no whitespace between them, but
there is one blank line before and after a set of namespace block closing lines.
To generate the namespace comments, I used clang-format, with
`FixNamespaceComments: true`. But since clang-format can't exactly duplicate
our astyle formatting, it made lots of other changes, too. To isolate the
namespace comments from the other formatting changes, I first ran clang-format
with `FixNamespaceComments: false`, committed those changes, then ran it again
to generate the namespace comments. Then I stashed the namespace comments,
reset `HEAD` to remove the other changes, applied the stashed namespace
comments, and committed the results (after examining them and making a few minor
adjustments by hand).
Signed-off-by: Michael Richters <gedankenexperimenter@gmail.com>
To make it easier - or in some cases, even possible - to put the keyboard into
programmable mode, we need to be able to initiate a reset from the host side.
For example, ynlike on AVR, where we can send a HUP to the serial port, on GD32,
we need to do it a bit differently: by sending the keyboard a Focus request to
reboot itself.
This implements the command itself.
Signed-off-by: Gergely Nagy <algernon@keyboard.io>
This implements a new `FocusSerial` command: `plugins`. The `plugins` command
will reply back with a list of plugins enabled in the firmware. The list is not
exhaustive, only plugins that opt-in to this mechanism will be listed. It is
opt-in, because for a lot of plugins, having them listed isn't useful in a
practical sense.
The goal with this feature is to allow Chrysalis to detect plugins that would
affect what keys it offers, or which additional settings it displays, and do so
in a consistent way. This is why IdleLEDs has an `onNameQuery` handler too, even
though it can be detected otherwise: for consistency.
Signed-off-by: Gergely Nagy <algernon@keyboard.io>
Gergely Nagy
Taylor Yu
Michael Richters
Jesse Vincent