Build system problem (was: mrkSoftTest doesn't compile)

[Ralph Lange <Ralph.Lange@bessy.de>]

The two -Wl options (compiled-in locations for shared libraries) are 
correct:
One location is the generic location, where the EPICS Base will be 
installed. The other is the local directory, i.e. the location of the 
Base that I'm currently linking against. This makes sure that the 
libraries are found on the development machine as well as on the runtime 
system (where the Base will be installed at the generic location).

There is one major problem, though:
The library path for the current linker command (-L) is set to the 
generic location!
That means, that even if I'm pointing my local configure/RELEASE to my 
base under test (/home/unix/lange/work/epics/APS/epics/base/3-14-X), any 
executables will be linked against the generic libraries in the 
.../3-14-7-0 location, which are the old ones.

I assume that - while keeping both locations in the -Wl seems correct - 
the library path for the linker command should always be the location 
directly from the EPICS_BASE location set in configure/RELEASE

Janet, would you agree?

Ralph


Marty Kraimer wrote:

> It works for me on my version of Linux. Mine, however, is getting old
>
> uranus% cat /proc/version
> Linux version 2.4.21-9.0.1.EL (bhcompile@bugs.devel.redhat.com) (gcc 
> version 3.2.3 20030502 (Red Hat Linux 3.2.3-26)) #1 Mon Feb 9 22:44:14 
> EST 2004
>
> In
> /usr/bin/g++ -o common  
> -L/opt/epics/R3.14.7/support/base/3-14-7-0/lib/linux-x86/         
> -Wl,-rpath,/opt/epics/R3.14.7/support/base/3-14-7-0/lib/linux-x86 
> -Wl,-rpath,/home/unix/lange/work/epics/APS/epics/base/3-14-X/lib/linux-x86 
>
>
> Why are there two "-W1" pointing to different versions of base?
>
> Marty
>
>
> Ralph Lange wrote:
>
> > Marty,
> >
> > mrkSoftTest doesn't compile aymore (neither on HP nor on Linux):
> >
> > aCC -o access  -L/opt/epics/R3.14.7/base/3-14-7-0/lib/hpux-parisc/   
> > -AA -mt       
> > -Wl,+b/home/unix/lange/work/epics/APS/epics/base/3-14-X/lib/hpux-parisc:/home/unix/lange/work/epics/APS/epics/extensions/lib/hpux-parisc:/opt/epics/R3.14.7/base/3-14-7-0/lib/hpux-parisc,+s       
> > access_registerRecordDeviceDriver.o asinittest.o asTest.o 
> > asTrapWriteTest.o accessMain.o    -lrecIoc -lsoftDevIoc -liocsh 
> > -lmiscIoc -lrsrvIoc -ldbtoolsIoc -lasIoc -ldbIoc -lregistryIoc 
> > -ldbStaticIoc -lca -lCom
> > /usr/ccs/bin/ld: Unsatisfied symbols:
> >   pvar_int_dbRecordsOnceOnly (first referenced in 
> > access_registerRecordDeviceDriver.o) (data)
> >
> > On Linux:
> >
> > /usr/bin/g++ -o common  
> > -L/opt/epics/R3.14.7/support/base/3-14-7-0/lib/linux-x86/         
> > -Wl,-rpath,/opt/epics/R3.14.7/support/base/3-14-7-0/lib/linux-x86 
> > -Wl,-rpath,/home/unix/lange/work/epics/APS/epics/base/3-14-X/lib/linux-x86 
> > -Wl,-rpath,/home/unix/lange/work/epics/APS/epics/extensions/lib/linux-x86        
> > devAiAsyn.o common_registerRecordDeviceDriver.o commonMain.o    
> > -ltestDevIoc -lrecIoc -lsoftDevIoc -liocsh -lmiscIoc -lrsrvIoc 
> > -ldbtoolsIoc -lasIoc -ldbIoc -lregistryIoc -ldbStaticIoc -lca -lCom
> > common_registerRecordDeviceDriver.o(.text+0x4a1): In function 
> > `__static_initialization_and_destruction_0':
> > /net/unix/mnt/csr-home/lange/work/epics/APS/epics/mrkSoftTest/common/src/O.linux-x86/common_registerRecordDeviceDriver.cpp:264: 
> > undefined reference to `pvar_int_dbRecordsOnceOnly'
> >
> > I don't see any changes in mrkSoftTest, though. Strange. I'm pretty 
> > sure you know faster than me what's wrong ...
> >
> > Ralph