When many channels are selected for recording a race condition can cause a crash by writing to a file that has not been opened yet.

It's unused, declared obsolete, and is triggering GCC warnings. If
anyone wants it back, it'll be in the Git history.

Signed-off-by: Marti Bolivar <mbolivar@leaflabs.com>

Signed-off-by: Marti Bolivar <mbolivar@leaflabs.com>

Signed-off-by: Marti Bolivar <mbolivar@leaflabs.com>

getInputChannelName() and getOutputChannelName() (pure virtual in the
superclass) aren't returning anything. So far this hasn't triggered
any memory errors because nobody's calling them.

To hack around this, have both of these return
"xxx-UNUSED-OPEN-EPHYS-xxx" (none of the existing subclases override
these methods, so leaving them pure virtual isn't workable). This way,
if we ever start calling these, we'll notice right away that a fix is
needed (instead of segfaulting, etc.).

Also keep them virtual to allow overrides.

Signed-off-by: Marti Bolivar <mbolivar@leaflabs.com>

That's not gonna work. Since nullParam is just a "nothing found", put
it in static storage and return that.

Signed-off-by: Marti Bolivar <mbolivar@leaflabs.com>

It looks like declaring them to return bool was a mistake:

1. None of their callers check their return value.

2. They don't return a value, and it doesn't look like there's any
   reason for them to.

Signed-off-by: Marti Bolivar <mbolivar@leaflabs.com>

If it's for internal use only, then enforce that.

Signed-off-by: Marti Bolivar <mbolivar@leaflabs.com>

Fixes caught by -Wreturn-type.

Have enable() properly return a value. Have disable() actually disable
the Arduino, and then properly return a value.

Signed-off-by: Marti Bolivar <mbolivar@leaflabs.com>

Signed-off-by: Marti Bolivar <mbolivar@leaflabs.com>

Generic changes:

- Genuinely unused variables (-Wunused-variable): delete
  them. Sometimes add a comment if the variable is for an extension of
  the code that isn't done yet.

- Variables that became unused due to commented-out
  code (-Wunused-variable): comment out their definitions. Dirty, but
  that's because commenting out code instead of deleting it is
  dirty.

  Many of these are due to commented-out printlines that should have
  been log messages that only get emitted at high verbosity level.

Specific changes:

- SpikeObject.cpp (-Wunused-value): move initialization of loop
  variable into first for loop clause to silence the warning.

Signed-off-by: Marti Bolivar <mbolivar@leaflabs.com>

Signed-off-by: Marti Bolivar <mbolivar@leaflabs.com>

Rearrange member initialization in constructors to silence GCC
-Wreorder warnings (-Wreorder is enabled by -Wall).

Signed-off-by: Marti Bolivar <mbolivar@leaflabs.com>

We still hardcode paths, but this is better.

- Initialize lengthOfInputFile and bufferSize to 0 in case input isn't
  found.

- Pull call to fopen out of ifdefs, so a future patch can just rip out
  the hardcoded path parts.

- If input _is_ found, then don't leak the file descriptor when the
  FileReaderThread gets deleted.

- If input isn't found, emit a sensible error message and bail in the
  constructor when the file isn't found. Have subsequent calls that
  try to use the FileReaderThread fail instead of crash.

Signed-off-by: Marti Bolivar <mbolivar@leaflabs.com>

Signed-off-by: Marti Bolivar <mbolivar@leaflabs.com>

Signed-off-by: Marti Bolivar <mbolivar@leaflabs.com>

There's no such thing (type, variable, or otherwise) as a FileReader.

Signed-off-by: Marti Bolivar <mbolivar@leaflabs.com>

Signed-off-by: Marti Bolivar <mbolivar@leaflabs.com>

Signed-off-by: Marti Bolivar <mbolivar@leaflabs.com>

Signed-off-by: Marti Bolivar <mbolivar@leaflabs.com>

- We'll need this functionality exposed to fix other bugs elsewhere,
  so add new SourceNode::tryEnablingEditor().

- Use isReady() instead of dereferencing dataThread directly, since it
  may be NULL.

- Tweak console output to more accurately describe what's happening.

Signed-off-by: Marti Bolivar <mbolivar@leaflabs.com>

Signed-off-by: Marti Bolivar <mbolivar@leaflabs.com>

Just include the superclass (DataThreads/DataThread.h) header in
SourceNode.h, and move the includes for the existing subclasses into
SourceNode.cpp. This will keep users of SourceNode from depending on
implementation details of the various subclasses.

Signed-off-by: Marti Bolivar <mbolivar@leaflabs.com>

Signed-off-by: Marti Bolivar <mbolivar@leaflabs.com>

createProcessorFromDescription() purports to only want one Intan demo
board at a time, but in fact bars multiple "File Reader" or "Custom
FPGA" subProcessorType values. We're adding hooks for a socket reader
here, and the current code would limit us to just one, which isn't
what we want.

Signed-off-by: Marti Bolivar <mbolivar@leaflabs.com>

Signed-off-by: Marti Bolivar <mbolivar@leaflabs.com>

Uses the parameters array to create XML data for parameter settings on
each channel when saving.

This only works for processors that use the parameters array, though I
updated LFP and Signal Generator to update that in parallel (LFP seems
to have an unrelated bug in dealing with channels that is causing some
problems). Eventually, all parameters should be used in the parameters
array, since it opens up a lot of possible holes when using default
values (they need to be specified twice), and once loading is
implemented. I've fixed some of the problems with parameters, so this
should be easier to do.

As part of this, I also fixed the implementation of
addParameterEditors. The old way of implementing a custom parameter
editor was to rewrite the virtual function, but because it was called
in the constructor, this didn't work (constructors use the type of the
parent). Instead, all editors now also pass a bool
"useDefaultParameterEditors" which is used by addParameterEditors() to
determine whether to do anything. I think this is the best way to do
it. It's initialized to default as true, though I had some trouble with
it implicitly taking the argument and so had to add true to most of the
editor functions (which is probably good practice anyway).

I also fixed some of the functions for generic processor to get
parameter index and parameter name.

Obviously, this is of considerable less use without functions for
loading parameters, but I figured it'd be better to make available for
testing with just the loading in case the changes to Parameters broke
anything I missed.