| Age | Commit message (Collapse) | Author |
|---|
|
Sometimes Arduino-based USB devices don't work because some hardware
(like KVM switches) gets confused by the CDC sub-devices.
This change makes it relatively easy to disable CDC at compiletime.
Disabling it of course means that the serial console won't work anymore,
so you need to use the reset button when flashing.
CDC_DISABLED is also used in ArduinoCore-samd for the same purpose.
based on
https://github.com/gdsports/usb-metamorph/tree/master/USBSerPassThruLine
See also https://github.com/NicoHood/HID/issues/225 and
https://github.com/arduino/Arduino/issues/6387 and
https://forum.arduino.cc/index.php?topic=545288.msg3717028#msg3717028
|
|
Improve reading ADC result
|
|
7.3.0-atmel3.6.1-arduino7 gcc fails to optimize separate reading from ADCL and ADCH. It produces additionally three eor commands or in some cases two mov commands in the assembly code (see discussion #344). These commands swap register contents before store them to data area. So they are completely unnecessary. Reading ADC result with ADC macro fixes it and gcc generates the right code..
|
|
The standard dictates that `std::size_t` is used, rather than the plain
`size_t` type.
Even though these types are usually, if not always, exactly the same
type, other code might assume that `std::size_t` is actually used and thus
also available under that name after including `<new>`.
This fixes that by using the right type. One challenge is that it is
usually declared in headers that we do not have available, so this just
defines the `std::size_t` type in the `<new>` header to work around
that.
|
|
This is currently disabled, keeping the old behavior of returning
NULL on failure, but should probably be enabled in the future as code
that does want to do a null check has had a chance to switch to the
more portable nothrow versions.
When enabled, allocation failure calls the weak `std::terminate()`,
which calls `abort()` by default, but can be replaced by user code to do
more specific handling.
To enable this, a macro must be defined (in new.cpp or on the compiler
commandline).
This fixes part of #287.
|
|
This fixes part of #287.
|
|
These are special functions that are presumably put into vtables for
deleted or pure virtual functions. Previously, this would call `abort()`
directly, but calling `std::terminate()` achieves the same effect, but
allows user code to change the behavior (e.g. to print to serial, blink
leds or whatever makes sense).
|
|
This allows calling it from other places later. The default
implementation calls `abort()`, but making it weak allows user code to
override this function (either directly, or by including a library like
uclibc++ that implements `std::set_terminate()`).
Note that this does not add a declaration for this function, since the
standard dictates this to be in `<exception>`, but we cannot
meaningfully or completely implement that header, so better leave it to
be overridden by e.g. libraries like uclibc++.
|
|
This makes these functions weak, so that a sketch or library can replace
them. This does not apply to all of these operators, only for the ones
that the C++ standard specifies as replaceable.
|
|
This makes this header complete up to including C++14, except two
exception classes that cannot be defined without `<exception>`.
The functions related to the "new_handler" are declared but not actually
defined, to prevent overhead and complexity. They are still declared to
allow implementing them in user code if needed.
This makes the implementation of all operator new and delete functions
comply with the C++11/C++14 specification in terms of which should be
actually implemented and which should be delegate to other functions.
There are still some areas where these implementations are not entirely
standards-compliant, which will be fixed in subsequent commits.
This fixes part of #287 and fixes #47.
|
|
Originally, the Arduino core used "new.h", rather than the standard
"new", probably because the implementation was incomplete, and for the
most commonly used new and delete operators, no include is needed at all
(they are defined implicitly by the compiler). However, now Arduino
does expose the "new" name, as an alias for the older "new.h". Given
that the standard name is "new", it makes more sense to put the actual
content in "new", and make "new.h" a compatibility header that includes
"new" instead of the other way around.
|
|
|
|
|
|
|
|
* Interrupt ordering for 32u2 and 16u2 MCU
* Added missing chip variants
|
|
No fixed value for USB power current.
|
|
Minor optimization in shiftOut function
|
|
Cast pins to signed integers to avoid Wtype-limits compile warning
|
|
Do not claim AT-protocol in CDC interface descriptor
|
|
|
|
Add placement new operator
|
|
Add parameter names to common prototypes
|
|
using with ternary operator such as bitWrite(value, bit, some_computed_value == 5 ? 1: 0);'
|
|
Fix unused variable warning for non-pluggable USB in SendDescriptor
|
|
checking `length` in below while statement
|
|
|
|
The CDC code presents itself as a virtual serial port. However, it also
sets the "bFunctionProtocol" value to 1, which means it supports
AT-commands, which is not actually the case. This might cause problems
with some software, such as ModemManager.
Originally, ModemManager would be very liberal with probing serial
devices, using a blacklist to prevent probing non-modems such as
Arduinos.
Since version 1.7.990, it has supported a "strict" mode where it tries to be
more restrained in what devices it probes. For CDC ACM devices, this
means it will only probe devices that claim to support AT-commands.
However, it also stopped applying the blacklist (intending to eventually
remove the blacklist), meaning it would again probe Arduinos.
This new strict policy is not the upstream default, but is enabled in
Debian (since Buster) and Ubuntu (since bionic 18.04.2).
The proper way to fix this, is to not claim AT comand support in the USB
device descriptor, which is what this commit does. The Arduino will
still show up as a virtual serial port, just not be probed by
ModemManager in strict mode.
For the commit that introduced the strict mode in ModemManager, see
https://cgit.freedesktop.org/ModemManager/ModemManager/commit/src?id=ee570d44dc117dc69f23e83313dd877f76c5e3e0
|
|
|
|
|
|
Avoids unused variable warning if USB is enabled but pluggable USB is not
|
|
Putting the parameter names in these commonly used prototypes makes syntax help like VS Code's Intellisense work 99% more useful. Without them, it doesn't give you the names of the parameters and you have to remember the semantics yourself. :(
|
|
These are currently implemented by the Wire library, on twi.c
|
|
Based on code originally by Rob van der Veer <rob.c.veer@gmail.com>, this adds
USBDevice.isSuspended(), so user sketches can run custom code in their `loop`
methods after checking if the device is suspended or not.
Signed-off-by: Gergely Nagy <algernon@keyboard.io>
|
|
The standard mandates that placement new should be have a noexcept specifier.
|
|
Casting to void is a well known trick for prevening 'unused parameter' warnings.
|
|
|
|
|
|
|
|
Make write to UDR and clearing of TXC bit in flush() atomic
to avoid race condition.
Fixes #3745 (second different issue introduced later but discussed
in the same issue)
|
|
Preserve values of configuration bits MPCMn and U2Xn.
Avoid setting other read-only bits for datasheet conformance.
See #3745
|
|
Moving the head buffer pointer and setting interrupt flag is now
atomic in write(). Previously an intervening ISR could empty the
buffer before the second ISR is triggered causing retransmission.
Fixes: #3745 (original issue only)
|
|
New macro TX_BUFFER_ATOMIC makes the following code block atomic
only if the transmit buffer is larger than 256 bytes. SREG is restored
on completion.
The macro is then used to simplify code for availableForWrite()
|
|
Instead of checking for the NEW_LUFA_SIGNATURE once in program memory and then
setting a flag which is used for further checks, a function is used that always
checks program memory directly.
If a flag is used, there's a slight chance that its location in RAM could fall
on MAGIC_KEY_POS. In this case, an aborted USB auto-reset sequence may fail.
|
|
An auto-reset invoked using USB CDC is triggered by the port closing
(when set to 1200 baud). Closing of the port is indicated by DTR going inactive.
There is no need to have auto-reset invoked by a CDC_SET_LINE_CODING command.
Only the CDC_SET_CONTROL_LINE_STATE command, which indicates a change in the
state of DTR, should be used.
|
|
The state of the watchdog timer is saved during a USB auto-reset and then
restored if the reset is aborted, in case the sketch is using the watchdog.
|
|
In the USB CDC code to invoke an auto-reset, the magic key location could be
restored before it had actually been saved. The sketch would then have a
corrupted value at this location. This fix prevents the value from being
restored if it hasn't previously been saved.
|
|
24, 44 and 84
|
|
|
|
These functions were changed from private to protected in https://github.com/arduino/Arduino/commit/99f2a2755349784835130147e46cb61659b85893 but the comments were not updated at that time.
In conjunction with equivalent pull requests to Arduino SAM Boards and Arduino SAMD Boards, solves https://github.com/arduino/Arduino/issues/6146.
|
|
see https://github.com/arduino/Arduino/pull/5789#discussion_r118007759
|
