When autoconsume is disabled even the basic example given in the wiki throws an exception. I've narrowed it down as much as possible, the following block:

!/usr/bin/env python

from gnuradio import gr
from gnuradio.extras import block_gateway
import numpy

class my_basic_pass_block(gr.block):
def init(self):
gr.block.init(self, name="my_basic_pass_block", in_sig=[numpy.float32], out_sig=[numpy.float32])
self.set_auto_consume(False)

def forecast(self, noutput_items, ninput_items_required):
ninput_items_required[0] = noutput_items

def work(self, input_items, output_items):
in0 = input_items[0]
out = output_items[0]
out[:] = in0
self.consume(0, len(in0))
return len(out)

Crashes with this message:

handler caught exception: operands could not be broadcast together with shapes (8192) (16383)
Traceback (most recent call last):
File "/usr/local/lib/python2.7/dist-packages/gnuradio/extras/block_gateway.py", line 53, in eval
try: self._callback()
File "/usr/local/lib/python2.7/dist-packages/gnuradio/extras/block_gateway.py", line 124, in __gr_block_handle
) for i in self.__out_indexes],
File "/home/user/mygnuradioblocks/my_basic_pass_block.py", line 18, in work
out[:] = in0
ValueError: operands could not be broadcast together with shapes (8192) (16383)
thread[thread-per-block[1]: <gr_block my_basic_pass_block (9)>]: caught unrecognized exception

But when commenting out the self.consume and self.set_auto_consume lines it works as expected.

The python functions inserted into gras::Block in block.i dont make it over into the share_ptr for some reason -- and we just manually copied them.

same idea as PMT manager:

• store allocated PMC objects
• use ref count to determine availability

(except this wont be a hack)

Yea yea, python is already JIT-tastic. This is really a convenience block to take a python source and to just be a GRC block. This saves the user time making xml blocks.

for users to experiment w/ feedback

• tags have standardized on a int64 + double time format, but dealing with this is a pain
• create an object to deal with time but w/ conversions from and to this format

with JIT so user can just add the code source at runtime

Currently we execute the clang binary to get llvm bitcode. It should be possible to use the clang api to directly compile code and get an llvm Module. Some work here: https://github.com/guruofquality/grextras/tree/use_clang_api

However, ubuntu 12.10 has a clangdev missing important libraries. 13.04 has the working dev package. So users dont have to manually install clang-dev from source... we dont want to force the api usage option.

So when I get around to upgrading my ubuntu -- do an ifdef to support both executing and API.

This block would take several time-tagged async inputs and use the tags to produce sample aligned outputs. This could align multiple usrp source blocks, and keep alignment in the face of packet overflows/dropouts. This block would be zero copy, implemented as a passive workflow block.

On the wiki, for the "Stream selector" block, the image is a broken link.

• accept tag on packet msg info
• post tag into output stream

• TCP/UDP
• client/server
• input and output

since we are using the c++ opencl api -- not sure

This breaks the python interface for the add/mult_const_v constructors.

Or it would be nice to use std::vectorstd::complex for all of them...

In any case, using std::complex isnt really heard of much.

-josh

I was building grextras on a 32-bit Ubuntu 10.04 machine, when the build tests failed:

tjt7a@bt:~/src/grextras/build$ ctest
Test project /home/tjt7a/src/grextras/build
Start 1: qa_add_and_friends
1/9 Test #1: qa_add_and_friends ..............._Failed 0.23 sec
Start 2: qa_block_gateway
2/9 Test #2: qa_block_gateway ................. Passed 0.24 sec
Start 3: qa_delay
3/9 Test #3: qa_delay ......................... Passed 0.19 sec
Start 4: qa_noise_source
4/9 Test #4: qa_noise_source .................._Failed 0.20 sec
Start 5: qa_signal_source
5/9 Test #5: qa_signal_source ................. Passed 0.20 sec
Start 6: qa_pmt
6/9 Test #6: qa_pmt ..........................._Failed 0.20 sec
Start 7: qa_msg_passing
7/9 Test #7: qa_msg_passing ..................._Failed 0.83 sec
Start 8: qa_pmt_to_python
8/9 Test #8: qa_pmt_to_python ................. Passed 0.20 sec
Start 9: qa_pmt_rpc
9/9 Test #9: qa_pmt_rpc ....................... Passed 2.21 sec

56% tests passed, 4 tests failed out of 9

Total Test time (real) = 4.51 sec

The following tests FAILED:
1 - qa_add_and_friends (Failed)
4 - qa_noise_source (Failed)
6 - qa_pmt (Failed)
7 - qa_msg_passing (Failed)
Errors while running CTest

I then looked more closely at the qa_msg_passing test and got the following:

tjt7a@bt:~/src/grextras/build$ ctest -VV -R qa_msg
UpdateCTestConfiguration from :/home/tjt7a/src/grextras/build/DartConfiguration.tcl
UpdateCTestConfiguration from :/home/tjt7a/src/grextras/build/DartConfiguration.tcl
Test project /home/tjt7a/src/grextras/build
Constructing a list of tests
Done constructing a list of tests
Checking test dependency graph...
test 7
Start 7: qa_msg_passing

7: Test command: /bin/sh /home/tjt7a/src/grextras/build/python/qa_msg_passing_test.sh
7: Test timeout computed to be: 9.99988e+06
7: linux; GNU C++ version 4.4.3; Boost_104000; UHD_003.005.001-release
7:
7: EEE
7: ======================================================================
7: ERROR: test_disconnect (main.test_msg_passing)
7: ----------------------------------------------------------------------
7: Traceback (most recent call last):
7: File "/home/tjt7a/src/grextras/python/qa_msg_passing.py", line 133, in test_disconnect
7: self.assertItemsEqual(sink.msgs(), msgs)
7: AttributeError: 'test_msg_passing' object has no attribute 'assertItemsEqual'
7:
7: ======================================================================
7: ERROR: test_hier (main.test_msg_passing)
7: ----------------------------------------------------------------------
7: Traceback (most recent call last):
7: File "/home/tjt7a/src/grextras/python/qa_msg_passing.py", line 103, in test_hier
7: self.assertItemsEqual(sink.msgs(), msgs)
7: AttributeError: 'test_msg_passing' object has no attribute 'assertItemsEqual'
7:
7: ======================================================================
7: ERROR: test_top (main.test_msg_passing)
7: ----------------------------------------------------------------------
7: Traceback (most recent call last):
7: File "/home/tjt7a/src/grextras/python/qa_msg_passing.py", line 84, in test_top
7: self.assertItemsEqual(sink.msgs(), msgs)
7: AttributeError: 'test_msg_passing' object has no attribute 'assertItemsEqual'
7:
7: ----------------------------------------------------------------------
7: Ran 3 tests in 0.314s
7:
7: FAILED (errors=3)
1/1 Test #7: qa_msg_passing ...................***Failed 0.84 sec

0% tests passed, 1 tests failed out of 1

Total Test time (real) = 0.84 sec

The following tests FAILED:
7 - qa_msg_passing (Failed)
Errors while running CTest

It looks like I'm missing "assertItemsEqual".

• using PMC data grams

• unit test has issue with the done logic - see comments
• need test for blocking, consuming, null/non producing source, etc

One of the things missing when we did precog was a programmatic way to access properties on the USRP. We sort of hacked around it with the PMT server calling into python functions in the top block -- its was a real kludge around and the code was a mess.

The properties interface in GRAS should make this seamless:
https://github.com/guruofquality/gras/wiki/Properties

What we need to do:

• make a uhd properties block that:
• opens a multi-usrp session
• registers handlers for all setters/getters
• register all known uhd swig types into PMC
• grc xml registers this block into tree using the $(id)

Given that, any block in the flow graph can access the USRP properties in a very clean and programmatic way; in c++ or python. Look how simple it can be:
https://github.com/guruofquality/gras/wiki/Codeguide#wiki-property-access

The C++ API was fun but it has build problems across versions, specifically with opnecl 1.2. The C API should be way less hassle.

the idea is to reduce breakage when updating and to minimize the changes needed in gr mainline

• the packet framer relies on a modified correlator block - copy this into the lib
• the noise source relies on gr random from core - copy this into lib

• this relies on the 0 item input reserve

• vector add/mult const
• as well as add, mult, sub, divide
• also subtract divide unit tests for single inputs

look for both types of devices, pick first one found

Hello gruel is no longer available since gnuradio 3.7 this causes cmake to fail on:
coudl not find module GRUEL

see also: online reference to problem

gr-extras doesn't work with gnu-radio 3.7: gruel and core are replaced by runtime as mentionned on this page (it may be helpful):
http://gnuradio.org/redmine/projects/gnuradio/wiki/Move_3-6_to_3-7
thanks

Take an input stream with tags, msgs, items and serialize it. This produces an output stream of type PacketMsg. The packet msg can be sent over TCP or even stored in a file.

Then vice-versa.

• for input and out messages

Would be nice to have initialization args or kwargs. Probably a PMCList and Dict. Also add this to the GRC XML wrapper.