Experimental Physics and
Industrial Control System

"David Dudley" <[email protected]> · Fri, 08 Sep 2006 11:25:47 -0500

Is QNX closer to Linux, or to BSD or Solaris?

In working on the NetBSD port, I started with the Solaris port, since NetBSD is a derivative of BSD4.4, which Solaris is also (distant relative, I understand).  Linux is fairly close, but not as close.

David Dudley

>>> David Eisert <[email protected]> 09/08/06 11:19 AM >>>
Andrew Johnson wrote:
> Hi David,
>
> Welcome to the EPICS community.
Hello Andrew,
Thank you. But actually I am on more of an unauthorized feasibility 
study. We get nagged by visitors from other facilities asking us 
"Where's the EPICS?". This is especially disturbing when it comes from 
members of review committees. I have somewhat of an uphill battle if I 
actually propose joining the EPICS community. If I can get base to build 
I may start another thread for help on this.
>
> You're going to have to port EPICS to QNX then, since we don't have 
> the operating system interface for that OS in Base. 
I was able to figure that out from the EPICS web site. Great job on 
documentation and training materials by the way.
>
> The part of the error message I'm interested in is this:
>
>> make[3]: *** [fdmgr.o] Error 1
>> make[3]: Leaving directory 
>> `/home/Epics/base-3.14.8.2/src/libCom/O.linux-x86'
>
> This tells me that the build failed while it was trying to compile 
> code that will run on Linux, not code that will run on QNX (which 
> would have been an error from inside some libCom/O.qnx-XXX 
> directory).  The EPICS build system was designed to compile multiple 
> architectures simultaneously alongside themselves, ...
I left out a bit too much information in my first post. QNX can be built 
from either a cross development system or native. In its most minimum 
configuration it can be flashed into a small SBC with only a MB or two. 
The full configuration requires more than a GB of hard disk space and 
will run X and larger open source packages like Eclipse, Qt and MySQL. I 
am running QNX as a native development system on a VME VMIC-7750 CPU 
board (733Mhz PIII /w 64MB RAM). Our IOCs are diskless but boot off a 64 
MB Compact Flash card. The parts of QNX that I need fill about half of 
the card. My development system is identical except that I have a small 
hard drive with a complete install of QNX. So from the EPICS build 
prospective it is almost identical to the Linux build environment. I 
have hijacked the Linux build scripts rather than potentially causing 
myself problems by creating a separate QNX build configuration for EPICS 
base. I realize I am editing files that I should not be editing but I am 
taking notes and if I can get EPICS to build I will untar a new copy of 
base and add proper QNX build configuration files. So everywhere you see 
O.linux-x86 just think O.qnx-x86. All of the code is being built to run 
under only one architecture.

PS: I have commented out the build of the code in the libCom/fdmgr 
directory. QNX complies the rest of libCom without error messages. It 
then proceeds to ToolComms and builds Flex and antelope with out error 
messages but bombs with a segmentation violation when it runs antelope. 
I am hoping that this is due to not being able to build the fdmgr code 
and not a new problem since antelope is being linked to libCom.

While I am waiting for feedback I am assembling the parts to run a newer 
version of QNX. The newer 6.3 version has gcc 3.3.5.

>
> Please post the output from the following command:
>     /usr/bin/g++ --version
> This should *not* be your QNX cross-compiler, it must be the compiler 
> for the host Linux operating system.  What Linux distribution are you 
> building this on?
>
> - Andrew

QNXDevel> /usr/bin/g++ --version
2.95.2

As mentioned above I am not running any copy of Linux. I have built base 
on a completely separate system running Fedora Core 5. But that is 
totally unrelated to this problem. Just an FYI that I have been able to 
build a working copy of base.

A bit of facility background
The facility is the Aladdin synchrotron storage ring in Madison, 
Wisconsin (SRC). We have about 20 IOCs with approximately 1000 physical 
control elements (our term is devices, yours is non-virtual PVs). Half 
of our IOCs are running PIII VMIC-7750 under QNX and the other half are 
running SBS VIPC-360 VME MC68360 using u/cos. The VIPC-360 are built 
into our dipole and quadrupole power supplies and have only a small 
portion of the total PVs. Our Consoles are Windows XP PCs so we do not 
have much experience with Unix systems. We used to be VAX/VMS based and 
we switched to Windows about ten years ago. The control system has all 
the major elements of EPICS but a lot fewer features. This control 
system has basically been my personal project for more than 20 years.

Thanks,

Dave

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System