Re: [Discuss-gnuradio] build error at uhd::time_spec_t::get_system_time()

Hello Michael,

yes, this is true:

# pacman -Ss libuhd

community/libuhd 3.10.2.0-3 [installed]

    Universal Software Radio Peripheral (USRP) userspace driver

So, apparently I had it installed. So, I did what you suggested (I think!):

git diff host/CMakeLists.txt 

diff --git a/host/CMakeLists.txt b/host/CMakeLists.txt

index ab7f41fb5..5e88d38c9 100644

--- a/host/CMakeLists.txt

+++ b/host/CMakeLists.txt

@@ -324,8 +324,8 @@ SET(Boost_ADDITIONAL_VERSIONS

 )

 FIND_PACKAGE(Boost ${BOOST_MIN_VERSION} COMPONENTS ${BOOST_REQUIRED_COMPONENTS})

 

-INCLUDE_DIRECTORIES(${Boost_INCLUDE_DIRS})

-LINK_DIRECTORIES(${Boost_LIBRARY_DIRS})

+#INCLUDE_DIRECTORIES(${Boost_INCLUDE_DIRS})

+#LINK_DIRECTORIES(${Boost_LIBRARY_DIRS})

 

 MESSAGE(STATUS "Boost include directories: ${Boost_INCLUDE_DIRS}")

 MESSAGE(STATUS "Boost library directories: ${Boost_LIBRARY_DIRS}")

Hope these are the correct lines to comment out. 

But now - since the documentation for PyBOMBS is not that clear (and it seems scattered among several places):

should I issue

pybombs rebuild uhd

If so - that did not work - I got exactly the same error. I may try to remove the uhd package with pacman and retry,

however I just learned from PyBOMBS guys that only Python 2 is supported, so I may now have a mess on my system.

Must check and clean it up before I continue.

Thanks for your help,

Peter

------

Piotr Swiatkiewicz

On 25 March 2018 at 18:51, Michael Dickens <michael.dickens@ettus.com> wrote:

Hi Piotr - The error you encountered generally means that UHD was already installed in the same prefix as the build would be going into, and that the UHD API changed between the versions. Normally UHD can be build over a pre-existing install, but sometime not too long ago this feature was broken via these 2 lines in the top level CMakeLists.txt file: < https://github.com/EttusResearch/uhd/blob/maint/host/CMakeLists.txt#L327 >. Specifically for the build part, the "INCLUDE_DIRECTORIES(${Boost_INCLUDE_DIRS})" comes before some (but not all) of the local-to-build INCLUDE_DIRECTORIES, which means that if the UHD API has changed, then the old API will be picked up -- as its headers come before the new (internal to build) headers. The easy fix for most OSs is to comment out those 2 lines & give the build a go again. I'm working on fixing this issue in a more robust way, since almost all of my builds / testing are on systems with UHD already installed into the build's prefix. Hope this helps! - MLD

On Sun, Mar 25, 2018, at 7:43 AM, Piotr Swiatkiewicz wrote:

I am trying to build GR with PyBOMBS and apart from problems with the build system itself (it would not start properly when python3 is used) I get the following error not far away (85% complete) from the end of the build:

[ 85%] Building CXX object gr-uhd/swig/CMakeFiles/_uhd_swig.dir/uhd_swigPYTHON_wrap.cxx.o

/home/pswiatki/devel/gr/default/src/gnuradio/build/gr-uhd/swig/uhd_swigPYTHON_wrap.cxx: In function 'PyObject* _wrap_time_spec_t_get_system_time(PyObject*, PyObject*)':

/home/pswiatki/devel/gr/default/src/gnuradio/build/gr-uhd/swig/uhd_swigPYTHON_wrap.cxx:19850:34: error: 'get_system_time' is not a member of 'uhd::time_spec_t'

       result = uhd::time_spec_t::get_system_time();

                                  ^~~~~~~~~~~~~~~

make[2]: *** [gr-uhd/swig/CMakeFiles/_uhd_swig.dir/build.make:71: gr-uhd/swig/CMakeFiles/_uhd_swig.dir/uhd_swigPYTHON_wrap.cxx.o] Error 1

make[1]: *** [CMakeFiles/Makefile2:13301: gr-uhd/swig/CMakeFiles/_uhd_swig.dir/all] Error 2

make: *** [Makefile:163: all] Error 2

PyBOMBS.Packager.source - ERROR - Build failed. See output above for error messages.

PyBOMBS.Packager.source - ERROR - Problem occurred while building package gnuradio:

Build failed.

PyBOMBS.install_manager - ERROR - Error installing package gnuradio. Aborting.

I tried building with python3 (3.6) and also with python2 (2.7). It fails at exactly the same point.

Interestingly, I read this at http://files.ettus.com/manual/classuhd_1_1time__spec__t.html:

static time_spec_t uhd::time_spec_t::get_system_time ( void  )

static

DEPRECATED – Will get removed from UHD in future versions.

Get the system time in time_spec_t format. Uses the highest precision clock available.

So, can it be it actually got removed and there is some incompatibility between UHD and GR?

By the way, ByBOMBS pulled these commits from the respective repositories:

UHD:

commit 4844f66dccaa71da102b02bba7b8caf8f84a932c (HEAD -> master, origin/master)

Author: Trung N Tran <trung.n.tran@ni.com>

Date:   Wed Mar 14 08:31:33 2018 -0700

GNU Radio:

commit 273392bda86688194df03cd1c3eef0819236b3b3 (HEAD -> master, origin/master, origin/HEAD)

Author: japm48 <japm48@users.noreply.github.com>

Date:   Thu Mar 15 14:37:40 2018 +0100

What am I doing wrong?

_______________________________________________

Discuss-gnuradio mailing list

Discuss-gnuradio@gnu.org

https://lists.gnu.org/mailman/listinfo/discuss-gnuradio