Experimental Physics and
Industrial Control System

Jeff Hill <[email protected]> · Wed, 06 Nov 1996 11:17:31 -0700

Guy Jennings wrote:
> 
> Epics folk:
> 
> I recently ran into a problem while attempting to use a sync group from
> a C++ client application.  The code below is based on acctst.c in the
> epics
> distribution.
> 
>         /* int ca_sg_create(const unsigned int *g); */
> 
>         int     status;
>         CA_SYNC_GID     gid=0;  /* const unsigned int gid = 0; */
> 
>         printf("Initially %ud\n", gid);
> 
>         status = ca_sg_create(&gid);
>         SEVCHK(status, NULL);
> 
>         printf("Then %ud\n", gid);
> 
> If compiled with gcc the code works fine, with the ca_sg_create call
> initialising
> 'gid' to a non-zero value (45 in my case).  If compiled with g++ the
> code does not
> work correctly and 'gid' retains the value 0 after the call to
> ca_sg_create.  This
> appears to be due to the definition of CA_SYNC_GID as 'const unsigned
> int'. I think
> it would be safer to remove the const from the typedef for CA_SYNC_GID
> since at
> present you are relying on undefined behavior - (typecasting away the
> const in
> ca_sg_create).
> 
> As far as I can tell, CA_SYNC_GID is the only epics-defined type which
> displays this
> problem and the header already allows for CA_SYNC_GID not to be 'const'
> 
> Guy Jennings
> BESSRC CAT
> Argonne National Laboratory

I also discovered this problem during testing about a week ago. This
appears to track with GNU C version 2.7.2 (I saw the bug when I compiled
with gcc 2.7.2 and I used the -O flag). I also arrived at the conclusion 
that the use of "const" in the typedef for CA_SYNC_GID only leads to
problems - so I removed it. This should not require any change in
any of the CA clients. I recently committed this change to 3.13 beta
at APS.

If there is interest I will patch 3.12.

I have only seen this problem show up when compiling with gnu C/C++ 2.7.2 
(which is not distributed with vxWorks 5.2). 

Jeff

-- 
______________________________________________________________________
Jeffrey O. Hill                 Internet        [email protected]
LANL MS H820                    Voice           505 665 1831
Los Alamos, NM 87545 USA        FAX             505 665 5107

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System