gr-audio: update comments in wavfile_sink and wavfile_source and add list of supported audio files #6503

This avoids a race condition which causes messages to be skipped while the flow graph finishes execution. Signed-off-by: Clayton Smith <argilo@gmail.com>

Signed-off-by: Josh Morman <jmorman@gnuradio.org>

Format *_c_pydoc_template.h and *_python.cc using clang-format to make the generated files clang_format compliant. It's better to use clang-format than to modify the mako templates. Using clang-format directly you have nothing to do when changing the formatting rules but only run the binding process again, but you must not change the mako templates. As with 956f332 pywidget() was replaced qwidget() generic_python_cc.mako had to be modified, to generate the correct binding code. Signed-off-by: Volker Schroer <3470424+dl1ksv@users.noreply.github.com>

Valgrind reports that poly_pack reads uninitialized memory, leading to undefined behaviour. This happens because it rounds the length up to the next multiple of 32 before reading the polynomial coefficients. Initializing the coefficients to zero solves the problem. Signed-off-by: Clayton Smith <argilo@gmail.com>

This fixes undefined behaviour (left shift of a negative number) reported by gcc's Undefined Behaviour Sanitizer. Signed-off-by: Clayton Smith <argilo@gmail.com>

Signed-off-by: Clayton Smith <argilo@gmail.com>

The qa_filterbank test occasionally fails, and from the error message it's apparent that on these occasions outdata2 contains the same values as outdata. This suggests that the while loop that waits for the new filter taps to kick in is exiting too soon. I suspect this occurs because it compares floating point values for exact equality. I've changed this to check for approximate equality instead. Signed-off-by: Clayton Smith <argilo@gmail.com>

When running QA tests, gcc's Undefined Behaviour Sanitizer reports signed integer overflows in fxpt.h. I've fixed them by explicitly casting the input to an unsigned integer prior to the addition step. Signed-off-by: Clayton Smith <argilo@gmail.com>

gcc's Undefined Behaviour Sanitizer reports load of a misaligned address. Replacing the offending code with a memcmp fixes the issue. Signed-off-by: Clayton Smith <argilo@gmail.com>

Signed-off-by: Clayton Smith <argilo@gmail.com>

In #5804 I accidentally caused a compiler warning: warning: suggest braces around initialization of subobject [-Wmissing-braces] This change fixes the warning while still initializing the polyout array to zero. Signed-off-by: Clayton Smith <argilo@gmail.com>

This brings crc16_async_bb in line with crc32_async_bb, making it easier to compare them. Signed-off-by: Clayton Smith <argilo@gmail.com>

The CRC16 and CRC32 async blocks can perform unaligned reads during verification, which results in undefined behaviour. To fix this, I've switched them to use memcmp. The CRC16 validation was also incorrect, because it read two extra bytes, resulting in verification failures. I've fixd that here as well. Signed-off-by: Clayton Smith <argilo@gmail.com>

Signed-off-by: Ron Economos <w6rz@comcast.net>

Signed-off-by: Clayton Smith <argilo@gmail.com>

In addition, Fedora 36 and Ubuntu 22.04 have changed the Python install scheme to put modules under <prefix>/local. Our default install location of /usr/local results in files being install under /usr/local/local. The change to GrPython.cmake puts default installs back in /usr/local. (Suggested by dl1ksv). Signed-off-by: Jeff Long <willcode4@gmail.com>

Signed-off-by: Clayton Smith <argilo@gmail.com>

This works around matplotlib/matplotlib#21998 by importing PyQt5 before matplotlib. The bug was introduced in matplotlib 3.5 and fixed in 3.5.2, but Ubuntu 22.04 ships with 3.5.1. Signed-off-by: Clayton Smith <argilo@gmail.com>

This prevents gr-perf-monitorx from incorrectly reporting connection failures when unrelated exceptions occur. Signed-off-by: Clayton Smith <argilo@gmail.com>

Signed-off-by: Clayton Smith <argilo@gmail.com>

Signed-off-by: wakass <wakass@users.noreply.github.com>

In a few instances, the format string does not match the types of the arguments. In other cases, the format string is not constant. I fixed both problems here, and verified that C++20 type checking is happy with the updated format strings. Signed-off-by: Clayton Smith <argilo@gmail.com>

These values should correctly be clipped to INT_MAX. Signed-off-by: Michael Roe <michael-roe@users.noreply.github.com>

gnuradio-runtime no longer depends on Boost::system or Boost::regex, so it should be safe to remove these from target_link_libraries. Signed-off-by: Clayton Smith <argilo@gmail.com>

Set cpp flag in low/high_pass_filter_taps block Replase firdes. by filter::firdes:: in all *_filter_taps blocks Switch to correct include file in filter_fir_filter_xxx block Enabling enable_legend and setting spectrum width in qtgui_freq_sink_x block Testing with gr-filter/examples/filter_taps.grc Signed-off-by: Volker Schroer <3470424+dl1ksv@users.noreply.github.com>

C++20 is stricter about template syntax, and does not allow redundant template parameter lists in constructors and destructors. I've fixed the two instances that occur in GNU Radio, and verified that compiling in C++20 mode succeeds. Signed-off-by: Clayton Smith <argilo@gmail.com>

Signed-off-by: Clayton Smith <argilo@gmail.com>

Build cpp strings correctly Signed-off-by: Volker Schroer <3470424+dl1ksv@users.noreply.github.com>

The previous implementation was applying the rotation to the wrong indexes when multiple scheduled phase increment updates were processed in the same work call. This patch fixes the input/output buffer indexes on the `rotateN()` call and covers the multi-update scenario on QA. The previous QA implementation was not covering this bug because the multi-update test case was only checking the tags produced on each phase increment update and not whether the actual IQ output was correct. Signed-off-by: Igor Freire <igor@blockstream.com>

* grc: Add parentheses on arithmetic expressions If a variable consists of an aritmetic expression an is used in another expression the expression should be put in parentheses Signed-off-by: Volker Schroer <3470424+dl1ksv@users.noreply.github.com> * Extend operator checking to * / Signed-off-by: Volker Schroer <3470424+dl1ksv@users.noreply.github.com>

Signed-off-by: Clayton Smith <argilo@gmail.com>

Signed-off-by: Ryan Volz <ryan.volz@gmail.com>

…nda-forge-pinning 2022.06.10.18.09.10 Signed-off-by: Ryan Volz <ryan.volz@gmail.com>

Signed-off-by: Ryan Volz <ryan.volz@gmail.com>

The gr-wavelet module has gnuradio-blocks in its target_link_libraries even though it doesn't use this module. Removing it will prevent extraneous libraries from being linked. Signed-off-by: Clayton Smith <argilo@gmail.com>

The gr-pdu module has gnuradio-blocks and gnuradio-filter in its target_link_libraries even though it doesn't use these modules. Removing them will prevent extraneous libraries from being linked. Signed-off-by: Clayton Smith <argilo@gmail.com>

In 30643f0 all the calls to get_initial_sptr were replaced by make_block_sptr. However, there are a few places where get_initial_sptr still remains. This replaces the reminaing get_initial_sptr calls. Signed-off-by: Daniel Estévez <daniel@destevez.net>

This pr adds a cpp template to the time_sink Signed-off-by: Volker Schroer <3470424+dl1ksv@users.noreply.github.com>

Previously, if either libiio or libad9361 were found as a framework bundle on macOS, linking would fail because it would try to link to the framework directory and not the actual library file. This change sets the found library as an interface library on the imported target, which lets CMake automatically handle the case of a framework versus a normal library file. Signed-off-by: Ryan Volz <ryan.volz@gmail.com>

When generating cpp code for the waterfall sink (float), the setting of the Spectrum Width ( Half, Full ) is ignored and always set to Full Signed-off-by: Volker Schroer <3470424+dl1ksv@users.noreply.github.com>

It appears this workaround was copied from the gr-wavelet module, but it's not needed here because gr-pdu doesn't depend on GSL. Removing it reduces the number of libraries that gr-pdu links to. Signed-off-by: Clayton Smith <argilo@gmail.com>

* blocks: Add symbolinterleaver block This commit adds a symbol interleaver to gr-blocks. The idea is to specify an interleaver sequence to be able to interleave and deinterleave vectors of constant size. For convenience, the deinterleaver sequence is always derived. Thus, it is not necessary to manually derive the corresponding deinterleaver sequence. See issue #3637 for a preliminary discussion. #3637 Signed-off-by: Johannes Demel <demel@ant.uni-bremen.de> * blocks: Add type specializations to symbolinterleaver All common GNU Radio port types are supported now, i.e. bb, cc, ff, ii, ss. More type definitions may be added easily. Signed-off-by: Johannes Demel <demel@ant.uni-bremen.de> * blocks: Add GRC definition for symbolinterleaver In order to use the symbolinterleaver in GRC, the appropriate block definition was added. Signed-off-by: Johannes Demel <demel@ant.uni-bremen.de> * blocks: Fix fmtlib include for conda build The conda CI fails to build GR with `#include <fmt/core.h>`. According to online discussions, the correct include is `#include <spdlog/fmt/fmt.h>`. There's some hope that this change makes all CI pass. Signed-off-by: Johannes Demel <demel@ant.uni-bremen.de> * blocks: Rename to blockinterleaver The blockinterleaver was introduced as symbolinterleaver. Following a suggestion to this PR. The block is now called blockinterleaver. Signed-off-by: Johannes Demel <demel@ant.uni-bremen.de> * blocks: Fix MSVC non-standard keyword handling Since MSVC doesn't conform to logical keywords `and` and `or` by default, we can't use them. The `/permissive-` flag might change that but breaks the build in other places. Signed-off-by: Johannes Demel <demel@ant.uni-bremen.de> * blocks: Add huge interleaver test Since the `blockinterleaver` calls `set_output_multiple` with `interleaver_indices.size()`, we need to make sure this configuration doesn't break things. Signed-off-by: Johannes Demel <demel@ant.uni-bremen.de> * blocks: Fix for missing keyword `not` is a C++ keyword that is not available by default in some compilers. Albeit, it is a part of the C++ standard. Signed-off-by: Johannes Demel <demel@ant.uni-bremen.de> * blocks: Add block interleaver example This new block lacked an example. In the process some issues where identified and fixed. Signed-off-by: Johannes Demel <demel@ant.uni-bremen.de> * blocks: Move blockinterleaving to kernel namespace The actual interleaver implementation is a kernel and thus should reside in the kernel namespace. Signed-off-by: Johannes Demel <demel@ant.uni-bremen.de>

New feature in GCC12, helping find reliance on coincidental initialization Signed-off-by: Marcus Müller <mmueller@gnuradio.org>

fmtlib makes it easier to read the outcome of formatted strings. This is another place where fmtlib may show its benefits. Signed-off-by: Johannes Demel <demel@ant.uni-bremen.de>

Signed-off-by: Josh Morman <jmorman@gnuradio.org>

Since cairo.ImageSurface is pixel-based, it only accepts integer dimensions. No need to round up – proper rounding and padding make this quite safe. Signed-off-by: Marcus Müller <mmueller@gnuradio.org>

Signed-off-by: Marcus Müller <mmueller@gnuradio.org>

Signed-off-by: danielnachun <daniel.nachun@gmail.com>

Docstring incorrectly recommended negative seed values. Signed-off-by: Brett Gottula <bgottula@gmail.com>

Wrong values in the GUI HINT field are correctly handled in the gui but lead to useless error messages in the terminal window Signed-off-by: Volker Schroer <3470424+dl1ksv@users.noreply.github.com>

Signed-off-by: Marcus Müller <mmueller@gnuradio.org>

Fixes #5985 Signed-off-by: Volker Schroer <3470424+dl1ksv@users.noreply.github.com>

The warning message text for negative payload lengths was the same as the message for too large payload lengths, and was incorrect. Signed-off-by: esrh <esrh@gatech.edu>

Type of `seed` in noise_source was inconsistent with fastnoise_source and with the random number generator classes. This makes everything consistent. Signed-off-by: Brett Gottula <bgottula@gmail.com>

While we do want to have an uint64_t seed API in C++, Pybind will not allow us to pass in negative ints that way. If we instead settle on a wrapper that's int64_t, we only get positive ints up to 2**63-1. To get up to 2**64-1, we need to also explicitly declare an uint64_t wrapper, and forbid automatic type conversion. While figuring this out, I needed to add a bit of debug logging to fastnoise_source; that does feel like a slight enhancement, so I'm leaving that in. Signed-off-by: Marcus Müller <mmueller@gnuradio.org>

Signed-off-by: AlexandraTrifan <Alexandra.Trifan@analog.com>

Signed-off-by: Josh Morman <jmorman@gnuradio.org>

In 6ff02a5, we changed the method of determining the installation directory to work with Fedora 36 and Ubuntu 22.04. For custom prefixes, this did not end up working on Ubuntu 22.04. Signed-off-by: Jeff Long <willcode4@gmail.com>

Signed-off-by: Josh Morman <jmorman@gnuradio.org>

* qtgui: Add output message ports to range widgets Signed-off-by: Charles Duffy <charles@dyfis.net> * qtgui: move gui_hint below showports Signed-off-by: Josh Morman <jmorman@gnuradio.org> Co-authored-by: Josh Morman <jmorman@gnuradio.org>

In the GR_PYBIND_MAKE_CHECK_HASH the include option is disabled. So the python bindings of the qt-gui modules can't be rebuild, as the the include directories of qt5 and qwt are not used at this point. Therefore in gr-qtgui/python/qtgui/bindings/CMakeLists.txt GR_PYBIND_MAKE is used instead of GR_PYBIND_MAKE_CHECK_HASH and the binding files *_python.cc have no header section. Now the include option is enabled in GR_PYBIND_MAKE_CHECK_HASH and uses the variable extra_includes. To be compatible with the actual calls to GR_PYBIND_MAKE_CHECK_HASH where no extra_includes are set, the argument for the include parameter in bind_intree_file.py now may be empty. Signed-off-by: Volker Schroer <3470424+dl1ksv@users.noreply.github.com>

The CMakeLists.txt file shows how to provide required include directories for qt5 and qwt, so that the files sink_c_pydoc_template.h and sink_c_python.cc can be regenerated if necessary. Signed-off-by: Volker Schroer <3470424+dl1ksv@users.noreply.github.com>

* cmake: make paths relative to gnuradio root Signed-off-by: William G Ledbetter <williamgledbetter@gmail.com> * cmake: change GNURADIO_XXX_DIR to PROJECT_XXX_DIR Signed-off-by: William G Ledbetter <williamgledbetter@gmail.com>

Previously, parsing signatures would fail if the io_signature call was not prefixed with gr::, which is not necessary if currently in the gr namespace anyway. This error may be encountered when copying gnuradio source files into a new module, which don't use the gr:: prefix. This patch addresses #5800. Signed-off-by: esrh <esrh@gatech.edu>

Signed-off-by: Josh Morman <jmorman@gnuradio.org>

Remove the section at the top and the top level header Signed-off-by: Josh Morman <jmorman@gnuradio.org>

Provide the invalid value alongside the accepted range to ease with debugging. Signed-off-by: Fabian P. Schmidt <kerel@mailbox.org>

Update STRIP_FROM_INC_PATH and EXCLUDE so that build paths are not part of the generated documentation. Signed-off-by: A. Maitland Bottoms <bottoms@debian.org>

* Fix various typos Found via `codespell -q 3 -L ans,busses,circularly,fo,hist,impres,inh,inout,ist,ith,merchantibility,nd,nin,numer,ro,shif,sinc,uint -S ./volk,CHANGELOG.md,./docs/RELEASE-NOTES*,/home/beast/Projects/repos/gnuradio/.packaging/debian/changelog,/home/beast/Projects/repos/gnuradio/.packaging/fedora/gnuradio.spec` Signed-off-by: luz paz <luzpaz@github.com> * Update the hash in the corresponding pybind11 bind file Signed-off-by: luz paz <luzpaz@github.com> Signed-off-by: luz paz <luzpaz@github.com>

Signed-off-by: Paul Atreides <maud.dib1984@gmail.com>

d_items_remaining was not being updated during seek, so end/repeat of playback would not happen at the correct time. Signed-off-by: Jeff Long <willcode4@gmail.com>

Also, add a small helper so that we can use settings from .clang-format Signed-off-by: Marcus Müller <mmueller@gnuradio.org>

Signed-off-by: Josh Morman <jmorman@gnuradio.org>

cmake-format -c .cmake-format.py \ -i cmake/cmake_uninstall.cmake.in cmake/Modules/* \ cmake/Toolchains/arm_cortex_a8_native.cmake \ cmake/Toolchains/oe-sdk_cross.cmake \ $(git ls-files | grep -x mako CMakeLists.txt ) Signed-off-by: Josh Morman <jmorman@gnuradio.org>

@mormj

libfmt-9.0.0 disabled automatic std::ostream insertion operator (operator<<) discovery when fmt/ostream.h is included to prevent ODR violations. It require explicit tagging of ARG types. Tried the following: fmt::streamed and fmt::ostream_formatter not compatible with older versions, FMT_DEPRECATED_OSTREAM will be removed in the next major release so can't use. With the help of @mormj, fixed this issue by not passing pointers of types not supported by libfmt into the logging functions. Bug: https://bugs.gentoo.org/858659 Closes: #6052 Signed-off-by: Huang Rui <vowstar@gmail.com> Signed-off-by: Huang Rui <vowstar@gmail.com>

xfce4-terminal is the terminal emulator for xfce users urxvt is a commonly used terminal emulator preferred by many over xterm Support for gnome and kde specific terminal emulators are already here, so please accept this as well. Signed-off-by: Rick Farina (Zero_Chaos) <zerochaos@gentoo.org> Signed-off-by: Rick Farina (Zero_Chaos) <zerochaos@gentoo.org>

- in the GRC yml file the widget was incorrectly stored in the variable - in the py.cmakein file a variable was misspelled and caused errors Signed-off-by: Miklos Maroti <mmaroti@gmail.com> Signed-off-by: Miklos Maroti <mmaroti@gmail.com>

The RX2 in-phase was returning the same samples as RX1 in-phase The index in function fmcomms2_source_impl() needs to be incremented by 2 Fix submitted by Nicolas Florio (github: nicolasflorio1) Signed-off-by: Jeff Long <willcode4@gmail.com>

…library Some tests use spdlog indirectly, but avoid linkage through the gnuradio::gnuradio-runtime target. Signed-off-by: Marcus Müller <mmueller@gnuradio.org>

These are functionally superceded by the generic crc_append and crc_check blocks and will be removed in a future release Signed-off-by: Josh Morman <jmorman@gnuradio.org>

`uhd_find_devices` would be the default shell function to discover available USRPs. PyUHD provides a function `uhd.find("")` that does the job. Besides, in `gr-uhd/python/uhd/__init__.py` there is a definition of `find_devices`. However, this function is broken at the moment because it tries to use a `find_devices_raw` function from the gr-uhd pybind11 interface that is not available. With this commit, we may be able to fix `find_devices` as well. The actual change is just copied from UHD `host/lib/device_python.cpp`. Alternative solutions would include a new function in C++ and a pybind11 wrapper. Signed-off-by: Johannes Demel <demel@ant.uni-bremen.de>

Tag the output stream with the frequency, sample rate and hardware time. Following gr-uhd's rx_freq, rx_rate and rx_time tags. Closes #6106 Signed-off-by: Roman Vaughan <nzsmartie@gmail.com>

Signed-off-by: Ryan Volz <ryan.volz@gmail.com>

Revealed in Windows build using Boost 1.78. Signed-off-by: Ryan Volz <ryan.volz@gmail.com>

Signed-off-by: Vasil Velichkov <vvvelichkov@gmail.com>

Signed-off-by: Ryan Volz <ryan.volz@gmail.com>

Signed-off-by: Josh Morman <jmorman@gnuradio.org>

Signed-off-by: Ryan Volz <ryan.volz@gmail.com>

Signed-off-by: Clayton Smith <argilo@gmail.com>

Signed-off-by: Sam <sam@whitings.org> Signed-off-by: Sam <sam@whitings.org> Co-authored-by: Sam <shwhiti@sandia.gov>

Signed-off-by: Josh Morman <jmorman@gnuradio.org>

…size_t) Signed-off-by: Gwenhael Goavec-Merou <gwenhael.goavec-merou@trabucayre.com>

Signed-off-by: Michael Roe <michael-roe@users.noreply.github.com> Signed-off-by: Michael Roe <michael-roe@users.noreply.github.com> Co-authored-by: Michael Roe <michael-roe@users.noreply.github.com>

This option causes a lot of extraneous libraries to be linked. It may no longer be needed. Signed-off-by: Clayton Smith <argilo@gmail.com>

- Adds loggers - Logs RFNoC connections made Signed-off-by: Martin Braun <martin.braun@ettus.com>

The RFNoC block impls had some missing overrides. Signed-off-by: Martin Braun <martin.braun@ettus.com>

This adds the set_/get_property APIs into rfnoc_block and uses it in the constructor to provide user settings. Signed-off-by: Martin Braun <martin.braun@ettus.com>

This exposes rfnoc_block_generic into Python land. This is useful for exposing RFNoC blocks that do not have explicit, C++-based block controllers in gr-uhd (or in out-of-tree modules). Signed-off-by: Martin Braun <martin.braun@ettus.com>

The GRC bindings for the TX/RX streamers now allow setting stream args. This allows passing stream arguments to UHD, e.g., for raw UDP streaming, or for setting spp values. Signed-off-by: Martin Braun <martin.braun@ettus.com>

- RX and TX streamers can now set vector lengths - The RX streamer GRC binding gets a 'byte' type Signed-off-by: Martin Braun <martin.braun@ettus.com>

This exposes uhd::rfnoc::node_t::set_properties() into GNU Radio space. It allows setting properties via string-based dictionary (string keys, string values). This commit also removes general_work() from the Python bindings, which was exposed accidentally via bindtool and should not be called from Python (it only throws an exception anyway). Signed-off-by: Martin Braun <martin.braun@ettus.com>

This adds gr::uhd::rfnoc_block::set_property and get_property to the Python API. The tricky bit is that these are templated functions. The Python versions therefore receive an optional, fourth argument 'type_hint' which loosely corresponds to the template argument in C++. If there is no type hint, we try and guess the correct type based on what came in from Python. However, that can be ambiguous (e.g., Python's integer could be an int, size_t, long long, uint32_t or whatever in C++). For these cases, we allow a string-based representation of the type. Signed-off-by: Martin Braun <martin.braun@ettus.com>

Signed-off-by: Martin Braun <martin.braun@ettus.com>

This adds two changes to the qtgui CMake: - The structure of lib/CMakeLists.txt now matches that of other in-tree modules. This is so that in the future, we can have optional sources, e.g. if we make QtOpenGL an optional dependency, and want to add blocks based on its existence. - We find the Qt5 package with QtWidgets as a COMPONENT, rather than directly for Qt5Widgets. This allows adding OPTIONAL_COMPONENTS later on. Signed-off-by: Martin Braun <martin.braun@ettus.com>

Original Python code concatenates list with a tuple, and that fails. This patch replaces list by a tuple, so now it works correctly. This fixes issue #6243 Signed-off-by: Roman Dobrodii <ztcoils@gmail.com>

Signed-off-by: Clayton Smith <argilo@gmail.com>

Signed-off-by: Sahil Siddiq <icegambit91@gmail.com> Update hash in hilbert_fc_python.cc

It appears that std::atomic does not work on all platforms. Switching to the mutex approach already sued by the vector version of this block avoids the problem. Signed-off-by: Clayton Smith <argilo@gmail.com>

Without this declaration, function calls that require a time_spec_t argument must be provided with such an argument, even if there is a default argument. For example, in the RFNoC DDC block, we couldn't call set_freq() as such: ```python ddc_block.set_freq(0.0, 0) ``` It would produce the following error: ``` TypeError: set_freq(): incompatible function arguments. The following argument types are supported: 1. (self: gnuradio.uhd.uhd_python.rfnoc_ddc, freq: float, chan: int, time: gnuradio.uhd.uhd_python.time_spec_t = 0.0) -> float ``` Even though the argument `time` is optional, the default value of 0.0 is not able to satisfy the requirements. By adding the `py::implicitly_convertible` call, we declare float convertible to time spec, and the call above can proceed. Signed-off-by: Martin Braun <martin.braun@ettus.com>

Signed-off-by: Clayton Smith <argilo@gmail.com>

This makes boost::asio look identical to non-Boost asio within our block definitions. Signed-off-by: japm48 <japm48gh@gmail.com> Co-authored-by: Martin Braun <martin.braun@ettus.com> Signed-off-by: Martin Braun <martin.braun@ettus.com>

Signed-off-by: Clayton Smith <argilo@gmail.com>

In one instance, the deleter was not specified when constructing a shared pointer to a matrix. I've added it here. Signed-off-by: Clayton Smith <argilo@gmail.com>

As noted in the documentation for NamedTemporaryFile, it is not possible on all platforms to open the temporary file a second time before closing it. Thus, we need to be sure to close the file before opening it in a flowgraph. Signed-off-by: Clayton Smith <argilo@gmail.com>

Signed-off-by: Clayton Smith <argilo@gmail.com>

Python bindings for ControlPort were not ported to pybind11, so these tests have been non-functional since GNU Radio 3.9. Signed-off-by: Clayton Smith <argilo@gmail.com>

Signed-off-by: Jeff Long <willcode4@gmail.com>

Signed-off-by: Josh Morman <jmorman@gnuradio.org>

Prior to this change, C32 vectors were inadvertently serialized as (imag, real) on little-endian platforms and (real, imag) on big-endian platforms. Signed-off-by: Clayton Smith <argilo@gmail.com>

GitHub Actions has a default timeout of 360 minutes. The formatting and Doxygen jobs only take about two minutes each, so we should lower the timeout to ensure that these jobs fail in a timely manner if a hang occurs. Signed-off-by: Clayton Smith <argilo@gmail.com>

Python's Wave_read.readframes() function returns samples in machine order, so they should be unpacked in machine order and not little-endian. Signed-off-by: Clayton Smith <argilo@gmail.com>

Signed-off-by: Clayton Smith <argilo@gmail.com>

The close() method in the File Sink block does not work, so it's necessary (on Windows, at least) to wait for the File Sink to be destroyed before the output file can be deleted. Signed-off-by: Clayton Smith <argilo@gmail.com>

Signed-off-by: Clayton Smith <argilo@gmail.com>

These tests fail on almost every OSX CI run, with readings as low as 4 on the second num_message() call, and as low as 12 on the third. Signed-off-by: Clayton Smith <argilo@gmail.com>

Signed-off-by: Clayton Smith <argilo@gmail.com>

Signed-off-by: Ryan Volz <ryan.volz@gmail.com>

Use updated CXXFLAGS that allows for the default (oldest) deployment target. Signed-off-by: Ryan Volz <ryan.volz@gmail.com>

Signed-off-by: Ryan Volz <ryan.volz@gmail.com>

…nda-forge-pinning 2022.11.02.18.55.19 Signed-off-by: Ryan Volz <ryan.volz@gmail.com>

Signed-off-by: Clayton Smith <argilo@gmail.com>

The add_field16, add_field32, and add_field64 functions have undefined behaviour because they copy the machine representation of multi-byte integers directly into the output buffer. As a result, they do not produce the correct output on big-endian systems. Switching to bit shifts will ensure consistent operation across platforms. Signed-off-by: Clayton Smith <argilo@gmail.com>

The current 20-bit hash is prone to collisions. Using the full hash avoids that problem. Signed-off-by: Clayton Smith <argilo@gmail.com>

Signed-off-by: Clayton Smith <argilo@gmail.com>

This avoids issues with register lengths varying by platform. Signed-off-by: Clayton Smith <argilo@gmail.com>

Signed-off-by: Clayton Smith <argilo@gmail.com>

Running on a 32-bit platform should help catch more floating-point and integer-length bugs. Signed-off-by: Clayton Smith <argilo@gmail.com>

This fixes the corresponding QA tests, which used to fail on big-endian platforms. Signed-off-by: Clayton Smith <argilo@gmail.com>

qa_fecapi_cc_buffer_overflow is skipped because qa_fecapi_cc is a substring of its test name. Tightening the test regex fixes tihs. qa_hier_block2 in gnuradio-runtime is skipped because it shares the same name as a test in gr-blocks. I've renamed the gr-blocks test to avoid the conflict. Signed-off-by: Clayton Smith <argilo@gmail.com>

Signed-off-by: Ron Economos <w6rz@comcast.net>

This will result in more detailed error messages, making it easier to diagnose test failures. Signed-off-by: Clayton Smith <argilo@gmail.com>

The python bindings for pll blocks did not have the base class inheritance for control loop, so these methods were not available as callbacks Signed-off-by: Josh Morman <jmorman@gnuradio.org>

Signed-off-by: Josh Morman <jmorman@gnuradio.org>

Signed-off-by: Jiří Pinkava <j-pi@seznam.cz>

This reverts commit 94d6ed9. Signed-off-by: Clayton Smith <argilo@gmail.com>

This reverts commit 0427b2a. Signed-off-by: Clayton Smith <argilo@gmail.com>

* Windows implementation had several bugs * inverted "is this the first items" logic * inverted "1/magnitude finite" logic * allocations/deallocations every work() * specifying alignment of std::vector header doesn't align data * non-windows / GNU implementation had bug * Variable-length arrays in C++ are GCC extension * both versions had several bugs, both hard and perf * thread-safety of setters of floating point numbers * first-run work code's numerical output values depend on workload size (but we're in output_multiple mode, so easy to fix) * counter-productive cache thrashing to improve vectorization at the expense of having VLA/runtime allocation * unnecessary division in comparison * AGC3 test on windows was disabled, which hid the above bugs By removing one unnecessary hot-loop division, at worst-case identical performance replace VOLK's invsqrt with a plain sqrt, by which I divide. Introduced mutable setter lock. Signed-off-by: Marcus Müller <mmueller@gnuradio.org>

This avoids an exception when the size of the widget reaches zero. Signed-off-by: Clayton Smith <argilo@gmail.com>

This eliminates the need to link gnuradio-runtime with Python and fixes segfaults on conda+macOS when trying to use any Python block. Those segfaults would happen because the Python library was loaded twice: once dynamically from linking with gnuradio-runtime, and once from the running Python interpreter which conda-forge links statically with libpython. The block_gateway was only used from its Python binding, so the implementation now just lives there. This is an API change! Signed-off-by: Ryan Volz <ryan.volz@gmail.com>

Signed-off-by: Ryan Volz <ryan.volz@gmail.com>

None of the declarations in the embed.h header are needed to compile these Python bindings. Signed-off-by: Ryan Volz <ryan.volz@gmail.com>

After commit 956f332 removing pyqtwidget in favor of only qtwidget, Python has not actually been needed to compile gr-qtgui (other than the Python bindings). This removes all remaining traces of Python, including including the Python.h header and linking with Python. Signed-off-by: Ryan Volz <ryan.volz@gmail.com>

Declaring a library as PUBLIC in target_link_libraries tells downstream linkers of the target to also need to link to the PUBLIC libraries. This is necessary when the headers of those libraries are required by the public headers of the target library. There were some instances where linked libraries were declared PUBLIC but were not required to use the public headers, so they have been switched to PRIVATE declarations. This will allow OOT modules to reduce their linking and dependencies in some cases. Signed-off-by: Ryan Volz <ryan.volz@gmail.com>

Signed-off-by: Marcus Müller <mmueller@gnuradio.org>

Now, python blocks can also log using self.logger.{trace,debug,info,warn,error,critical}("string") Subtasks: * Adding log level setters/getters to Python interface of logging system * Removal of misleading commentary * unit tests Signed-off-by: Marcus Müller <mmueller@gnuradio.org>

Rationale: the average, especially the new user, benefits from logging. Required some overly talkative blocks to calm down. Signed-off-by: Marcus Müller <mmueller@gnuradio.org>

…g handler Signed-off-by: Marcus Müller <mmueller@gnuradio.org>

The connection flowgraph element previously kept a reference to the cairo context passed to the draw function in order to be able to use cairo's `in_stroke` function to determine if the mouse cursor was on the path of the curved connection line. As it turns out, this is dangerous because GTK is constantly destroying and creating new cairo contexts and surfaces. This avoids keeping a reference to the cairo context by initializing a local cairo context with the connection class for the sole purpose of storing the curved path and calculating `in_stroke`. The local context and surface can be very basic because not much is needed for `in_stroke`, and the context can persist and just have its path replaced when it needs to be updated. On Windows, resizing the GRC window (and particularly the flowgraph canvas) to a larger size than its initial size would cause the underlying cairo surface to be destroyed and a new one created. The cairo context stored for the curved connection line had a reference to that destroyed surface, and closing that context (upon replacing it with a new one) would attempt decrement the reference count on the surface. This would produce an assertion failure that crashed GRC stating `Assertion failed: CAIRO_REFERENCE_COUNT_HAS_REFERENCE (&surface->ref_count)`. On macOS with the Quartz backend, a related crash and assertion error would manifest when connecting blocks. Signed-off-by: Ryan Volz <ryan.volz@gmail.com>

All this did was request a larger size for the cairo surface of the DrawingArea on which this is set, which very roughly worked around the cairo context reference bug for users with high DPI displays. Now that this bug is fixed by the previous commit, this is not necessary. It's important to note that GTK takes care of the DPI scaling for high DPI displays behind the scenes, so this doesn't even help to scale the flowgraph properly. If there are further bugs with scaling, don't look here. Signed-off-by: Ryan Volz <ryan.volz@gmail.com>

Signed-off-by: Martin Braun <martin.braun@ettus.com>

Signed-off-by: Martin Braun <martin@gnuradio.org>

This puts gr-uhd at feature-parity with gr-ettus for the following blocks: FIR filter, siggen, and window. It also updates an issue in rfnoc_block regarding setting of properties. The FIR filter has different features starting in UHD 4.2. This is checked with a compiler flag. GRC Examples for those new blocks are also updated. Signed-off-by: Martin Braun <martin.braun@ettus.com>

Signed-off-by: Martin Braun <martin.braun@ettus.com>

When an RFNoC block is requested but not found, it now prints the exact block it was looking for in the exception string. Signed-off-by: Martin Braun <martin.braun@ettus.com>

Includes the following changes: - Add a set_streamer_adapter_id() API call to rfnoc_graph. This sets an adapter ID for later calls to connect(). - Amend the GRC bindings for the RX and TX streamers to allow specifying the adapter ID. It will call rfnoc_graph::set_streamer_adapter_id() to do so in the autogenerated Python code. Signed-off-by: Martin Braun <martin.braun@ettus.com>

In GRC, there was an issue where you could select a clock/time source for the RFNoC graph, but it was never passed on to the graph object. This fixes the issue such that the correct device args are generated. Signed-off-by: Martin Braun <martin.braun@ettus.com>

In ea8b4b6, the GRC bindings were added with an incorrect key. Signed-off-by: Martin Braun <martin@gnuradio.org>

Signed-off-by: Clayton Smith <argilo@gmail.com>

Signed-off-by: Jeff Long <willcode4@gmail.com>

Signed-off-by: Josh Morman <jmorman@gnuradio.org>

to reflect added fields up to v14 Signed-off-by: Josh Morman <jmorman@gnuradio.org>

Signed-off-by: Clayton Smith <argilo@gmail.com>

This Python function was accidentally broken when converting from SWIG to Pybind11. In 9c64b6d, uhd.find() was created as a replacement, but that leaves uhd.find_devices() in a broken state. This performes two changes: - The binding for uhd.find() is moved into python_bindings.cc from usrp_block_python.cc. - The Python version of find_devices() now correctly uses uhd.find() under the hood. - The SWIG version differentiated between Python- and C++-versions of device_addr_t and remapped that, which we can skip with Pybind11 Signed-off-by: Martin Braun <martin@gnuradio.org>

Signed-off-by: Clayton Smith <argilo@gmail.com>

Signed-off-by: Ron Economos <w6rz@comcast.net>

Signed-off-by: Clayton Smith <argilo@gmail.com>

The bug was fixed in Thrift 0.10.0, released in 2017. https://issues.apache.org/jira/browse/THRIFT-3932 Signed-off-by: Clayton Smith <argilo@gmail.com>

In Thrift 0.10.0, the byte type was deprecated and replaced with i8. Signed-off-by: Clayton Smith <argilo@gmail.com>

Signed-off-by: Clayton Smith <argilo@gmail.com>

Change gr::digital::ofdm_frame_equalizer_vcvc block to pass the tags to the equalizer that it aggregates. The gr::digital::ofdm_equalizer_base interface allows passing the tags vector to equalize() but in current implementation the default argument is used (empty vector). Fixes #6422 Signed-off-by: Mihai Stef <mihai.stef@com.utcluj.ro>

np.float as an alias for float has been deprecated and removed, and np.float_ or np.float64 is undoubtedly what was wanted here anyway. Signed-off-by: Ryan Volz <ryan.volz@gmail.com>

Signed-off-by: Josh Morman <jmorman@gnuradio.org>

Signed-off-by: ph1t0 <fito@ektocomms.space>

Allow new snippet insertion point in the init function but before blocks are instantiated. This allows actions or objects that might be required in the constructors of the blocks, such as connecting to a server, to be initialized Signed-off-by: Josh Morman <jmorman@gnuradio.org>

Importing gtk before requiring a specific version can cause issues if multiple versions of gtk are installed on a system. Signed-off-by: Seth Hitefield <sdhitefield@gmail.com>

Signed-off-by: Seth Hitefield <sdhitefield@gmail.com>

Signed-off-by: Jeff Long <willcode4@gmail.com>

Signed-off-by: Josh Morman <jmorman@gnuradio.org>

I have described the issue here #6459 Signed-off-by: Fran Boric <fboric@yahoo.com>

Signed-off-by: theartful <aessam.dahy@gmail.com>

Signed-off-by: Jeff Long <willcode4@gmail.com>

Signed-off-by: Grant Meyerhoff <grant.meyerhoff@ni.com>

Signed-off-by: Grant Meyerhoff <grant.meyerhoff@ni.com> Co-authored-by: Martin Braun <martin.braun@ettus.com> Co-authored-by: Grant Meyerhoff <grant.meyerhoff@ni.com>

bytes_read, the result of asio::ip::udp::socket::receive_from(), was an int, but the function returns size_t. This resulted in an overflow of the int and a subsequent comparison. Since the int result of buffer space_available() should never be negative, it is safe to cast the result to size_t. If that result is ever changed to something larger than size_t, this could cause trouble, but that seems unlikely. Signed-off-by: Jeff Long <willcode4@gmail.com>

3eea105 causes the gr-soapy block to add the rx_freq tag when set_frequency() is called with a name parameter. However, gr-soapy source's cmd port handler calls set_frequency() without name, so add_tag_ is not set and the rx_freq tag is not added. This patch causes rx_freq to be added for both versions of set_frequency(). Signed-off-by: Josh Bailey <josh@vandervecken.com>

This fixes a bug in the calculation of the frame size used by the tagged decoder that affects the CC decoder in terminated mode. Currently, tagged_decoder calls the set_frame_size() method of the decoder with a size computed as the input size multiplied by the rate. In the cc_decoder, the frame_size parameter of the set_frame_size() method is used to calculate and set d_veclen. The value of d_veclen is used to determine how many forward steps (butterflies) the Viterbi algorithm should be run. The number of input items that are read in d_veclen forward steps is d_veclen * d_rate (note that d_rate is actually the inverse of the rate; typically, d_rate == 2). Therefore, the size of the input buffer should be as large as this value, and in most cases it should be equal, because we want the cc_decoder to look at all the provided input. The d_veclen variable is related to d_frame_size, which indicates how many output bits are produced by the decoder. The value of d_frame_size is equal to the parameter of set_frame_size() in all cases except in terminated mode with byte padding, in which case the byte padding is subtracted when computing d_frame_size. The value of d_veclen is equal to d_frame_size + d_k - 1 in terminated and streaming modes, and to d_frame_size in truncated mode. The problem with the current code is that in tagged_decoder the set_frame_size() method is called with ninput_items[0] * rate. In the streaming and terminated modes this will cause d_veclen to be too large. The forward steps (butterflies) of the Viterbi algorithm will read past the end of the input buffer. This is undefined behaviour and can cause some bit errors at the end of the decoded codeword (the effect of reading past the end of the buffer is essentially extending the codeword with 12 symbols of garbage). This problem does not happen in the async_decoder because there is some code that computes a "diff", which works out to be d_k - 1 + d_padding * rate() in terminated mode and 0 in all other cases. The diff is subtracted from the frame size. This subtraction gives the correct d_veclen calculation, so that the butterflies do not read past the input end. This commit fixes the tagged_decoder when used with the terminated CC decoder by adding the same code as in the async_decoder to calculate and subtract this "diff" value. Some caveats: 1) This commit changes the size of the output packets of tagged_decoder when used with a terminated CC decoder. Formerly, the output packets were 6 bits longer, as if the tail bits were included in the output. However, having these 6 extra bits was just an artifact of looking 12 symbols past the end of the input. The cc_decoder_impl::chainback_viterbi() function is not prepared to extract tail bits in the terminated case. This chainback extracts bits "as they exit the shift register on the right", so tail bits cannot be extracted, because they "never exit the shift register". (Here I have in mind the usual depiction of a convolutional encoder, where input bits are shifted from left to right into a shift register). The CC_TRUNCATED case in cc_decoder_impl::generic_work() has some additional code to extract additional bits which haven't "exited the shift register" yet, because it is mandatory to extract them in this case (they are the last 6 bits of the message). It think there is no good use case for attempting to extract the tail bits in the terminated case, so this modification seems a good way forward. The new behaviour also matches that of async_decoder, which does not output tail bits. However, this breaks tag propagation of the tagged_decoder when used with the terminated CC decoder, because the block calls set_relative_rate() with the nominal rate (1/2), which is slightly different from the true rate once we take into account the fact that tail bits are dropped. This could be fixed by propagating tags manually in the work() function. 2) The streaming mode is still broken, both when used with the async_decoder and with the tagged_decoder. Since for the streaming case diff is 0 but d_veclen includes the term d_k - 1, the streaming mode reads past the input buffer. It isn't really possible to run the streaming mode properly with the async_decoder or tagged_decoder (it should be run with the decoder_impl.cc), because the streaming mode requires history (some overlapping input items between consecutive calls), which is not available with the async_decoder or tagged_decoder. 3) The code copied from async_decoder has a "watch out" comment stating that it might be over-specializing for the CC decoder in terminated mode. It is true that this code seems specially crafted to cover this very specific case, but as far as I have been able to think, it does not affect negatively any other cases (taking into account other modes of the CC decoder and other FEC decoders). Signed-off-by: Daniel Estévez <daniel@destevez.net>

Signed-off-by: R-ohit-B-isht <rbtunes0@gmail.com>

all these seem copy file-and-modify oversights; looks like no harm was done. Signed-off-by: Marcus Müller <mmueller@gnuradio.org>

Not required: e36f2fa ("pdu: Remove unused dependencies", 2022-06-01) Signed-off-by: Chris Mayo <aklhfex@gmail.com>

Signed-off-by: R-ohit-B-isht <rbtunes0@gmail.com>

…list of supported audio files Signed-off-by: harshitwr <singhh198@gmail.com> Signed-off-by: harshitwr <singhh198@gmail.com>

…vfile_documentation

…list of supported audio files. Signed-off-by: Harshit Singh <singhh198@gmail.com> Signed-off-by: harshitwr <singhh198@gmail.com>

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

gr-audio: update comments in wavfile_sink and wavfile_source and add list of supported audio files #6503

gr-audio: update comments in wavfile_sink and wavfile_source and add list of supported audio files #6503

Commits on Feb 2, 2023

Commits on Feb 3, 2023