Re: [gpsd-dev] gpsd Build Error "stdatomic.h" Not Found; Ubuntu 14.04 + ARM Cortex A-15

[Gr Hudy]

Cheers, Gary-

>> I didn't actually "mess with" any include directories or paths; I merely switched the cpp being used from cpp-4.8 to cpp-6, to match the version of gcc.

>Maybe not messing with the include directories directly, but certainly indirectly when you are flipping between cpp.  gcc should do that automagically, so if you touched that I'd double check what you did.

Sorry; my bad. Switching the default cpp doesn't actually change anything. As you pointed out, gcc automagically uses the correct cpp.

>Why don't you just turn off HAVE_STDATOMIC_H?  Comment out lines 783 to  794 in SConstruct.  You lose a bit of reliability in SHM export and time hinting.

In the copy I have (gpsd-3.17), the tests for std_atomic are lines 764 - 775 of SConstruct. I commented them out, and the build fails with:

g++-6 -o qt-gpsutils.os -c -pthread -Wextra -Wall -Wno-uninitialized -Wno-missing-field-initializers -Wcast-align -Wmissing-declarations -Wpointer-arith -Wreturn-type -O2 -pthread -fPIC -DHAVE_SYS_TIMEPPS_H=1 -DUSE_QT -DQT_SHARED -I/usr/include/dbus-1.0 -I/usr/lib/arm-linux-gnueabihf/dbus-1.0/include -I/usr/include/qt4 -I/usr/include/qt4/QtNetwork -I/usr/include/qt4/QtCore gpsutils.c

In file included from libgps.h:15:0,

                 from gpsutils.c:23:

compiler.h: In function 'void memory_barrier()':

compiler.h:93:5: error: 'atomic_thread_fence' is not a member of 'std'

     std::atomic_thread_fence(std::memory_order_seq_cst);

     ^~~

compiler.h:93:30: error: 'memory_order_seq_cst' is not a member of 'std'

     std::atomic_thread_fence(std::memory_order_seq_cst);

                              ^~~

scons: *** [qt-gpsutils.os] Error 1

scons: building terminated because of errors.

-Gr

----- Original message -----

From: "Gary E. Miller" <address@hidden>

To: address@hidden

Subject: Re: [gpsd-dev] gpsd Build Error "stdatomic.h" Not Found; Ubuntu 14.04 + ARM Cortex A-15

Date: Thu, 5 Jul 2018 13:59:18 -0700

Yo Gr!

On Thu, 05 Jul 2018 14:31:39 -0600

Gr Hudy <address@hidden> wrote:

> >Ouch.  I gave up on my Tk1 a long time ago.  Way too many problems

> >with it. 

> Sometimes I think that would be nice, but our current system is

> dependent on it. We make extensive use of the GPU, and run on

> batteries where the low power consumption is important. We are

> working on a new version that uses the Jetson TX2...

Let me know how it works.  For now I've given up on all the clones

and just stick to RasPi.

> >Which reminds me, I asked, and you did not answer, where is your

> >stdatomic.h? 

> D'oh! I had a niggling feeling I had forgotten something.

> /refsrc/gpsd-3.17 $ find / -name stdatomic.h

> /usr/lib/gcc/arm-linux-gnueabihf/6/include/stdatomic.h

> /usr/lib/gcc/arm-linux-gnueabihf/5/include/stdatomic.h

> /usr/lib/gcc-snapshot/lib/gcc/arm-linux-gnueabihf/4.9.0/include/stdatomic.h

> /refsrc/gcc-6.4.0/gcc/ginclude/stdatomic.h

>

> If I correctly read the output of the cpp command I used in the

> Jetson_Tk1_info.txt file I attached before, it should be using the

> one in /usr/lib/gcc/arm-linux-gnueabihf/6/include/stdatomic.h

Yes.  gcc 6 should use that one.

> I didn't actually "mess with" any include directories or paths; I

> merely switched the cpp being used from cpp-4.8 to cpp-6, to match

> the version of gcc.

Maybe not messing with the include directories directly, but certainly

indirectly when you are flipping between cpp.  gcc should do that

automagically, so if you touched that I'd double check what you did.

> I did this by slaving cpp off of gcc in

> update-alternatives, then selecting gcc-6 as the default compiler.

'update-alternatives' is a Debian thing, not a GNU thing.  I don't know

how that works.

Why don't you just turn off HAVE_STDATOMIC_H?  Comment out lines 783 to

794 in SConstruct.  You lose a bit of reliability in SHM export and

time hinting.

RGDS

GARY

---------------------------------------------------------------------------

Gary E. Miller Rellim 109 NW Wilmington Ave., Suite E, Bend, OR 97703

address@hidden  Tel:+1 541 382 8588

    Veritas liberabit vos. -- Quid est veritas?

    "If you can’t measure it, you can’t improve it." - Lord Kelvin

Email had 1 attachment:

+ Attachment2

  1k (application/pgp-signature)