Cleaned up RobotBase, removed singleton list from SensorBase, and removed unused typedefs and NULL_TASK macro from HAL's Task.hpp. Making the robot class instance static fixed non-POD statics used by the instance during destruction from being destroyed first.
As discussed, the old method of was going to have issues, and was not
going to help in the case of the structs containing pointers. I think we
are just going to have to be careful, as guarenteeing the internal
pointers are const is going to be very difficult.
As discussed, we don't really need the error checking specifically, as
we are just going to throw exceptions anyway, so this will make sure you
can check for 0 and be sure you catch any handle errors.
HAL Port is using a special handle, where the module and pin are bit
shifted straight into the handle. This is one of the few special cases
we have, but for the way port is used it is much cleaner and uses much
less memory. Plus it is generic and not specific to one type.
Internally, the linked list now uses shared_ptrs instead of raw
pointers. In addition, in the WPILib the notifier handle is now made
atomic. Then before the class is destructed, the handle is now set to 0.
This should help solve one of the existing race conditions. A 0 handle
is correctly handled down at the HAL level.
Handles will be replacing opaque pointers in the HAL for better compatibility and easier error checking. This commits adds the initial resource manangement and definition for handles.
In case of Spontaneous wake ups, we should be checking a condition
variable as well.
Note that we can not use the existing m_newControlData, as that has a
possible race condition with existing user code that it does not look
like we could work around.
The network communication headers were moved to the include directory of the HAL and their copies were removed from wpilibc. Also, symModuleLink.h was removed because the function which it prototypes hasn't been used in WPILib since the roboRIO switch.
Removed delayTicks(), delayMillis(), delaySeconds(), HAL_NO_WAIT, HAL_WAIT_FOREVER,
niTimestamp32(), and niTimestamp64().
Replaced clock_gettime() and usleep() with std::chrono.
Split to match the new headers. Uses a namespace 'hal' for internal functions and globals.
SPIAccumulator merged back into SPI header, as it was not a good split. Analog
accumulator will move back to analog input when the analog split is done.
* Replaced include guards with #pragma once
* All source files now have exactly one newline appended
Some files had either two newlines at the end or none (which isn't POSIX compliant). This patch fixes that.
Subsections are alphabetized according to lexographic ordering. Also, HAL includes were moved from headers to source files where possible. This change may cause user code which uses HAL functionality and does not include the relevant HAL header (since it may have been provided by another WPILib header) to fail to compile.
This is a follow-up of 3cd1253. A C++ program was written to automate the license update originally. That program was translated to Python so it can be kept in the repository and run when needed. It has been tested on Windows using the standard Python 3 installation and on Linux.
The original version skipped files that had "//" at the beginning since most were files that should be excluded. The relevant files are now in an exclusion list and the rest are processed normally. The .hpp file extension has been added as well. The script rewrote CompressorJNI.cpp to remove the carriage returns from its line endings.
The Digital and Analog headers (and the implementations, but that will
be moved over later) are just too big and congested. This splits those
headers, and then changes the few things that needed to be changed in
WPILib to get the code working again. No function changes were made in
this commit.
CANJaguar is the only motor controller using sync groups, so that feature doesn't belong in an interface used by all motor controllers. If teams want to use sync groups, they should cast to the appropriate motor controller type themselves.
As the notifierMutex is unlocked while the callback is being called in
alarmCallback(), it's possible for next to be invalidated due to a
cleanNotifier() call. Now, instead of deleting the notifier immediately,
add it to a freelist to be cleaned up at the tail end of alarmCallback.
