For usb_serial_jtag REPL only, xTaskCreate was passing a pointer to
esp_console_repl_com_t, while esp_console_repl_task was expecting
a pointer to esp_console_repl_universal_t.
The way the two structures are defined, this makes no difference, and
the pointer values are the same. Still, this could potentially break
in the future.
(I am not sure what is the distinction between repl_com (common?) and
repl_universal; it seems that `int uart_channel` could just as well
be part of esp_console_repl_com_t; alternatively, as suggested in the
previous commit, this structure could contain a callback function
pointer, which would allow `esp_console_new_repl_*` functions to
specify how stdin/stdout should be initialized by the REPL task.)
The crash occurred when calling setvbuf(stdin,...) with stdin==NULL.
This happened because esp_console_repl_task started running before
its args->uart_channel was initialized; then esp_console_repl_task
went into the code path 'uart_channel != CONFIG_ESP_CONSOLE_UART_NUM',
and tried to 'fopen("/dev/uart/0");`
Since the UART VFS is not registered when ESP_CONSOLE_USB_SERIAL_JTAG
option is enabled, fopen failed and 'stdin' was NULL.
Fix by moving the initialization of repl task arguments before the
start of the task, same as it is done for the usb_cdcacm case.
The crash started happening after the commit 287ab7566b. I haven’t
verified this, but I guess the reason why it wasn’t happening before
was that xTaskCreate was not correctly yielding to the newly created
higher-priority 'repl' task, therefore the code which was setting
the repl task arguments after xTaskCreate had time to execute.
It should be noted that the 'uart_channel' argument is a bit hacky,
in the first place. The code should be refactored to pass a callback
function to the repl task, and let this callback initialize stdin and
stdout based on the chosen console channel. Then esp_console_repl_task
does not require assumptions about the specific interface used.
Closes https://github.com/espressif/esp-idf/issues/9662
The actual output from the build tool (CMake/Ninja) may or may not
contain color escape codes, depending on various factors. The output
written to the log file should never include color escape codes,
though. This is because color escape codes in files are usually not
rendered as "color" in editors, and complicate reading. Also escape
codes would break the regular expressions used to display hints for
compilation errors.
If stdout is a TTY (meaning that the output is not redirected), tell
the build tool (GNU Make or Ninja) to enable colorized output.
GNU Make and Ninja also check if their stdout is redirected and
strip color escape sequences in that case. CLICOLOR_FORCE environment
variable overrides this behavior.
With this change, if the compiler was launched with the
-fcolor-diagnostics flag and idf.py output is not redirected, the
final output in the terminal will be colorized.
(-fcolor-diagnostics is handled at CMake level by the previous commit)
Related to https://github.com/espressif/esp-idf/issues/4162
Setting this option informs CMake that it should pass
-fcolor-diagnostics flag to the compiler.
(Colorized build system output, like from GNU Make, is produced even
without this flag.)
Note that if the build is done using Ninja and the build output is
redirected (not a TTY), Ninja will still strip the escape codes from
the output. For the case of idf.py, this is handled in the next
commit.
There are multiple changes in this commit:
1. Unify the RISC-V and ULP-FSM code paths in esp32ulp_mapgen.py.
It seems that these were originally introduced because `nm` output
for the RISC-V case contained symbol sizes, while for the ULP-FSM
no symbol sizes were reported. This makes sense, because the
ULP-FSM object files are produced from assembly source, symbol
sizes have to be added manually using the .size directive.
In the case of RISC-V, the object files are built from C sources
and the sizes are automatically added by the compiler.
Now 'posix' output format is used for both RISC-V and ULP-FSM.
2. Move BASE_ADDR out of esp32ulp_mapgen.py. This now has to be passed
from CMake, which should make it easier to modify if a new chip
with a different RTC RAM base address is added.
3. Add C++ guards to the generated header file.
4. Switch from optparse to argparse for similarity with other IDF
tools.
5. Add type annotations.
* "dummy loop to force pre-processed linker file generation" seems to
be unnecessary. It looks like the idea was copied from the
dependency of ULP-FSM preprocessed source files on the LD script.
* Can use add_dependencies instead of
set_target_properties(...LINK_DEPENDS...) which is more readable
* Use target_link_options instead of target_link_libraries, which is
supported starting from CMake 3.13. Unlike target_link_libraries,
it doesn't require manually quoting the pats.
This fixes the issue with build output not being colorized on Windows,
while the hints messages are colorized.
The issue occurred because sys.stdout and sys.stderr get overridden
by colorama.init() at runtime, but the default argument
output_stream=sys.stdout holds the reference to the"original"
sys.stdout.
colorama.init() (which, by the way, gets called via a curious chain
of imports, via idf_component_tools.manifest and tqdm package)
overrides standard streams, on Windows only. The overridden streams
contain logic to convert ANSI color codes into Windows Console API
calls to colorize the text.
Since read_and_write_stream function used the default value of
output_stream evaluated at module loading time, it was using the
original sys.stdout, not the one overridden by colorama.
One extra note is that while this does fix the coloring issue, the
solution is a bit fragile, as it relies on one of the following
(on Windows):
- colorama.init() is called (this can change if idf-component-manager
stops importing tqdm)
- Sufficiently new version of Windows 10 is used, and ANSI color codes
support is enabled in the Registry.
The actual output from the build tool (CMake/Ninja) may or may not
contain color escape codes, depending on various factors. The output
written to the log file should never include color escape codes,
though. This is because color escape codes in files are usually not
rendered as "color" in editors, and complicate reading. Also escape
codes would break the regular expressions used to display hints for
compilation errors.
If stdout is a TTY (meaning that the output is not redirected), tell
the build tool (GNU Make or Ninja) to enable colorized output.
GNU Make and Ninja also check if their stdout is redirected and
strip color escape sequences in that case. CLICOLOR_FORCE environment
variable overrides this behavior.
With this change, if the compiler was launched with the
-fcolor-diagnostics flag and idf.py output is not redirected, the
final output in the terminal will be colorized.
(-fcolor-diagnostics is handled at CMake level by the previous commit)
Related to https://github.com/espressif/esp-idf/issues/4162
Setting this option informs CMake that it should pass
-fcolor-diagnostics flag to the compiler.
(Colorized build system output, like from GNU Make, is produced even
without this flag.)
Note that if the build is done using Ninja and the build output is
redirected (not a TTY), Ninja will still strip the escape codes from
the output. For the case of idf.py, this is handled in the next
commit.
On the build runners, the installation frequently takes around 150
seconds, which is above the current timeout. This change increases
the timeout.
Closes IDFCI-1436
'linux' component provides some of the common header files, such as
'sys/queue.h' and 'sys/lock.h'. For chip targets, it is possible to
include these files without having to add any extra requirements.
With this change, the same behavior will apply for the linux target.
libpython2.7 was added to the container to allow running GDB built
with Python 2.7 support and distributed as part of the cross-compiler
toolchain.
Now that we have a new release of GDB which works with Python 3.x,
the GDB shipped with the cross-compiler is no longer used. Removing
libpython2.7 should reduce the image size.
This reverts commit be0372b1db.
spi_flash driver knows how to write data from cache (DROM or PSRAM)
into flash, so the extra check in nvs_flash is unnecessary.
Besides, the hardcoded address limit (0x3ff00000) is wrong for some
of the newer chips.
`__COMPONENT_TARGETS` is evaluated very early when components and
component directories are added to the build, which means that all
components (including the ones which are in EXCLUDE_COMPONENTS) have
a build system target defined. The component manager was given the
list of all known components (derived from the list of targets), not
the list of components after EXCLUDE_COMPONENTS were processed.
Because of that, EXCLUDE_COMPONENTS didn't effectively exclude the
component from the consideration of the component manager.
Changes accumulated since 20210826 release:
- Fixed cache flush emulation
- Fixed loading of baremetal ELF files via -kernel arg
- Added emulation of AES peripheral
- Fixed division by zero in UART emulation
- Fixed APP CPU using same ROM code as PRO CPU
- Base QEMU version upgraded from 6.1.0 to 7.0.0
Since both chips have built-in JTAG functionality, and there are no
official boards with FT2232H for these chips, use the built-in JTAG
by default.
To use them with esp-prog, set:
OPENOCD_COMMANDS="-f board/esp32c3-ftdi.cfg"
or pass this via the --openocd-commands argument to idf.py.
Blocking read from cdcacm VFS could return less bytes than requested.
This didn’t match the behaviour of other VFS drivers, and higher level
code could misbehave.
Previously, reset over USB CDC was done by calling esp_restart from
an interrupt handler. This works only until some restart hook function
is registered using esp_register_shutdown_handler, and the hook
function tries to do something that isn’t allowed in an interrupt
handler. One such case is with Wi-Fi. When Wi-Fi driver is installed,
it registers esp_wifi_stop as a shutdown handler function. However
esp_wifi_stop cannot be called from an ISR, and hence we shouldn’t
call esp_restart from an ISR either.
This commit modifies USB CDC driver to call esp_restart by posting it
to esp_timer task.
Closes https://github.com/espressif/esp-idf/issues/7404
This allows calling esp_log_level_set function from the console.
Together with CONFIG_LOG_MAXIMUM_LEVEL option, this allows enabling
logs from different tags at run time.
* Some components have initialization dependencies. To account for
them, simple numeric priority values are introduced.
* esp_system_init_fn_array moved into Flash from DRAM
* System init functions defined using ESP_SYSTEM_INIT_FN now return
an error code. This enables simpler and more consistent error
handling in the init functions. Returning an error from an init
function is now a valid approach — the startup code will print
an error and abort.
Many of the more complicated CMake projects can't be added to the IDF
build system simply by calling add_subdirectory.
"add_subdirectory" also cannot be used with projects which use build
systems other than CMake (for example GNU Make or cargo).
This commit changes the example to use ExternalProject_Add, instead,
which is a more general way of adding subprojects.
As part of this change, tinyxml2 is now downloaded from the Internet,
which allows removing one submodule.
XTAL_CLK_FREQ now depends on the actual XTAL used, remove this macro
and get the XTAL frequency from the RTC register instead.
No uses of XTAL_CLK_FREQ found, other than in the UART LL.
When ESP32-C2 is paired with a 26 MHz XTAL, the systimer tick
frequency becomes equal to 26 / 2.5 = 10.4 MHz. Previously we always
assumed that systimer tick frequency is integer (and 1 MHz * power of
two, above that!).
This commit introduces a new LL macro, SYSTIMER_LL_TICKS_PER_US_DIV.
It should be set in such a way that:
1. SYSTIMER_LL_TICKS_PER_US / SYSTIMER_LL_TICKS_PER_US_DIV equals the
actual systimer tick frequency,
2. and SYSTIMER_LL_TICKS_PER_US is integer.
For ESP32-C2 this means that SYSTIMER_LL_TICKS_PER_US = 52 and
SYSTIMER_LL_TICKS_PER_US_DIV = 5.
This introduced two possible issues:
1. Overflow when multiplying systimer counter by 5
- Should not be an issue, since systimer counter is 52-bit, so
counter * 5 is no more than 55-bit.
2. The code needs to perform:
- divide by 5: when converting from microseconds to ticks
- divide by 52: when converting from ticks to microseconds
The latter potentially introduces a performance issue for the
esp_timer_get_time function.
Some esp32c2 boards will be produced with a 26 MHz XTAL. This commit
adds the basic Kconfig option for this type of hardware.
Support for CONFIG_ESP32C2_XTAL_FREQ_26 in other areas of IDF will be
implemented in subsequent commits.
In C++20, using the result of an assignment to a 'volatile' value is
deprecated.
Breaking change: register "setter" or modification macros can no
longer be used as expressions.
Closes https://github.com/espressif/esp-idf/issues/9170
The bundled version of asio uses legacy concepts syntax
(https://en.cppreference.com/w/cpp/experimental/constraints) which is
not accepted by GCC with -std=c++20.
Disable the concepts support until asio is upgraded to a version which
supports the new syntax.
Ivan Grokhotkov