Experimental Physics and
Industrial Control System

"Ernest L. Williams Jr." <[email protected]> · Thu, 03 Nov 2005 09:10:19 -0500

On Thu, 2005-11-03 at 08:02 -0600, Marty Kraimer wrote:
> email message attachment (Returned mail: see transcript for details)
> On Thu, 2005-11-03 at 08:02 -0600, Marty Kraimer wrote:
> > The original message was received at Thu, 3 Nov 2005 06:41:26 -0600 (CST)
> > from mercury.aps.anl.gov [164.54.8.12]
> > 
> >    ----- The following addresses had permanent fatal errors -----
> > <[email protected]>
> >     (reason: 550 5.1.1 <[email protected]>... User unknown)
> > 
> >    ----- Transcript of session follows -----
> > ... while talking to apsmailhub.aps.anl.gov:
> > >>> DATA
> > <<< 550 5.1.1 <[email protected]>... User unknown
> > 550 5.1.1 <[email protected]>... User unknown
> > <<< 503 5.0.0 Need RCPT (recipient)
> > email message attachment
> > On Thu, 2005-11-03 at 08:02 -0600, Marty Kraimer wrote:
> > > Andrew Johnson wrote:
> > > 
> > > >
> > > >
> > > > I see 64 bit as a WIBNIF at present but not something we should delay 
> > > > R3.14.8 for.  Since we've already checked the config files in they 
> > > > should stay, but we don't advise use in production systems or for 
> > > > anyone who isn't prepared to do some debugging.
> > > >
> > > > - Andrew
> > > 
> > > Lets see if I can summarize the 64 bit issue.
> > > 
> > > When solarissparc and linux are built as "native 64". The following is true
> > > int  is 32 bits
> > > long is 64 bits
> > > u nsigned is 32 bits
> > > 
> > > This they are both the same.
> > > 
> > > exampleApp works on linux64
> > > exampleApp crashes on solaris
> > > 
> > > I traced the crash to db_access.c
> > > 
> > >     case(oldDBR_TIME_DOUBLE):
> > >         {
> > >                 struct dbr_time_double *pold=(struct dbr_time_double 
> > > *)pbuffer;
> > >                 struct {
> > >                         DBRstatus
> > >                         DBRtime
> > >                 } new;
> > >                 dbr_double_t *pvalue = &pold->value;
> > > 
> > >                 options=DBR_STATUS | DBR_TIME;
> > >                 nRequest=0;
> > > printf("before dbGetField pbuffer %p pold %p\n",pbuffer,pold);
> > >                 status = 
> > > dbGetField(paddr,DBR_DOUBLE,&new,&options,&nRequest,
> > >                         pfl);
> > > printf("after dbGetField pold %p\n",pold);
> > > 
> > > before dbGetField pbuffer 1001aa110 pold 1001aa110
> > > after dbGetField pold 0
> > > 
> > > 
> > > Note that pold, which is on the stack, was overwritten.
> > > 
> > > Without a lot of work, base is not ready for native 64 applications.

Marty, which compiler are you using for solaris-sparc64?

> > > 
> > > Marty
> > > 

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System