Re: question about aSubRecord

[wantianmin <wantianmin@gmail.com>]

Hi,Tim Mooney,
   Thanks a lot for your help,
   But when I try to use dbPutLink(aSubRecord* prec) to do my write 
operation, some compile error occurs.After I include all the include 
file need in aSubRecord.c ,the compiler complain about:

/usr/local/epics/base/include/dbFldTypes.h:26: error: expected 
identifier before numeric constant

 I have tried for some time today but failed.It seems difficult to use 
dbPutLink() to solution my problem ,and by reading your mail and the 
aSubRecord.c file I know the output link may be writtten  2 times or 
more ,which is not expected. So  I should use something others to do 
this , or this question may caused  by some unclear design. Some 
complicated operation may should not be done in aSubRecord.

Thanks for your help ,
Best Regards
                                                                                                         wantianmin
                                                                                                          SSRF

ä 2010å12æ02æ 23:11, Tim Mooney åé:
> wantianmin,
>
> It's possible to overcome this, though it might not be a good idea,
> depending on what link processing you have in mind.
>
> If you look at the record's process() routine, in
> base/src/rec/aSubRecord.c, you'll see what it's doing to execute
> output links.  If you do this in your routine, you can cause both
> input and output links to execute in any order you want.  If your routine
> returns 0, however, the output links will all be executed again.  If your
> routine returns other than 0 (no doubt this value was intended to signal
> an error), the record will not execute the output links, but (currently)
> this doesn't put the record into an alarm state, so the only side effect
> will be that the record ends up with a nonzero VAL field.
>
>      Maybe there should be a return value that calls for no output link
>      processing but that doesn't also indicate an error.
>
> If you plan an extended series of (write, read, decide, write, read, ...),
> or if any of the records you write to are asynchronous, and you wait for
> the result of their processing, your record is not going to play well with
> other records that are executed by the same task. (For example, if your
> record is periodically scanned, all the other records scanned with the same
> period will be forced to wait for your record to finish before they will be
> processed.)
>
> It might be that your purpose would be better served by an SNL program.
>
> Tim Mooney
>
> ----- Original Message -----
> From: "wantianmin"<wantianmin@gmail.com>
> To: tech-talk@aps.anl.gov
> Sent: Wednesday, December 1, 2010 8:55:28 PM
> Subject: question about aSubRecord
>
> Hi ,all:
> When I use aSubRecord record to do some process, It was found that write
> operation is not being done until myAsubProcess() finish.Is there some
> way to overcome it?
> here is a description of my situation:
> db:
> record(aSub,"asub")
> {
> field(INAM,"myAsubInit")
> field(SNAM,"myAsubProcess")
> field(FTA,"DOUBLE")
> field(NOA,"1")
> field(INPA,"test4 CA")
> field(FTVA,"DOUBLE")
> field(NOVA,"1")
> field(OUTA,"test1 CA")
> field(FTVB,"DOUBLE")
> field(NOVB,"1")
> field(OUTB,"test2 CA ")
> }
> dbSubExample.cpp:
> static long myAsubProcess(aSubRecord *precord)
> {
> if (mySubDebug)
> printf("Record %s called myAsubProcess(%p)\n",
> precord->name, (void*) precord);
> printf(" begin process\n");
> double* p1=(double*)precord->vala;
> p1[0]=20;
> printf(" write vala 1 to 20 complete....\n");
>
> double* p2=(double*)precord->a;
> printf(" read from a :%lf\n complete",p2);
>
>
> printf(" end process\n");
> return 0;
> }
>
> I found that When I doing dbpf asub.PROC 1 in iocshell,
> the first output value (test1) is not changed until the process finished.
> but in some application ,I need to write some pv and read it again to
> decide what is need to do next.
> Is there some suggestions?
> Thanks a lot
> Best Regards
> wantianmin
> SSRF