Re: unexpanded macros in comments

[Andrew Johnson <anj@aps.anl.gov>]

Hi Gasper,

Gasper Pajor wrote:
> I don't expect to find a lot of new EPICS issues these days, so I didn't 
> request a mantis account, although it would be a proper way to file a 
> bug :)

Actually I prefer to get reports by email first, so I can do some triage 
before they go into Mantis.  However, either way works.

> Anyway, when starting an IOC (epics 3.14.x on linux) with a template 
> file that has unexpanded macros in comments (i.e. in lines beginning 
> with '#'), the errors are printed to IOC console:
>
> ## Load record instances
> dbLoadRecords("db/dbExample1.db","user=gpajor")
> macLib: macro somemacro is undefined (expanding string # $(somemacro)
> )
> filename="../dbLexRoutines.c" line number=317
> macExpandString failed for file db/dbExample1.db
>
> I think the macro expanding mechanism should disregard the comment lines.

Unfortunately this is not currently possible because of the order in 
which we parse the .db files - macro expansion takes place before the 
text gets fed to the tokenizer and parser, so when we're looking up the 
macro values we don't know whether we're currently parsing a comment 
line or not.

There are advantages and disadvantages in this, and unfortunately I 
don't know whether anybody is making use of the addtional capabilities 
this provides. For example, this record definition and dbLoadRecords 
line are legal as far as the IOC's parser are concerned:

record($(type),"$(name)"){
  field($(field))
}

dbLoadRecords("db/example.db","type=ai,name=input,field='DTYP, xy240'")

I suspect that no tool like VDCT would accept the .db file, but if my 
understanding is correct the IOC would be perfectly happy with it.  It 
would even be possible to start a comment inside a macro value, although 
that would be a rather strange thing to do.

I'm not saying I particularly like the way this has been implemented, 
but that's what the code we currently have does, and before I can change 
it I would have to poll the community to see whether anybody is relying 
on it.

Any comments from unconventional users of macros like the above record 
definition would be appreciated, in case we decide to change things in 
the future.


There is a way to prevent the error message associated with an undefined 
macro, which is to give the macro a default value.  The following 
comment line should not generate any errors from dbLoadRecords:

# $(macro=)

The equals sign is used to introduce a value to be used in the event 
that the macro doesn't exist at the time it's expanded, and this may be 
sufficient for your purposes.  This syntax has been supported by macLib 
since Base R3.14.6.

- Andrew
--
The right to be heard does not automatically include
the right to be taken seriously. -- Hubert H. Humphrey