Re: Parsing idl ...

[Owen Taylor <otaylor redhat com>]

Michael Meeks <michael@helixcode.com> writes:

> Hi Christian,
> 
> On Tue, 27 Jun 2000, Christian Borup wrote:
> > Michael Meeks wrote:
> > > I forgot to say; the method used in Owen's perl ORBit bindings is
> > > particularly evil; since libIDL forks a preprocssor (gpp) to chew the idl
> > > beforehand, thus making it rather unsuitable for use on a non-development
> > > machine.

The user of cpp on non-development machines is not uncommon - people
have used it for things like preprocessing their .Xresource files.
Red Hat, for one, does ship a separate cpp package for this reason.  

> > It shouldn't be messing with the idl in the first place and only does so
> > because an IR is not always available. I think there is some problems with
> > the ORBit IR, which is what it should be using.
> 
> 	There may well be problems with the IR, but Owen assures me IIRC
> that parsing the idl is a rather quicker method than using the IR ( and
> this would seem to be obvious over a network connection ) and that ORBit
> perl is sticking with it. Either way, I tend to think the flexibility of
> the IR approach is far greater.

The biggest problem with an IR is keeping it properly fed with the
correct up-to-date files. I found this to be a huge pain when I was
developing using MICO's IR. It's less of a problem with a deployed
system, but if you are changing your IDL, an IR is very inconvienient
compared to just reading in the IDL files from disk when the
client or server is restarted.

ORBit Perl does include most of the code to work with an IR, but
it hasn't been maintained or tested recently because once I decided
to write the IDL parsing code and got it running, it was just a lot
nicer to use than running from an IR. (I used to have scripts
that would kill the IR and refill it from the correct IDL files.) 

If the ORBit IR is finished (I haven't tracked it recently),
someone writes the necessary code to fill the ORBit IR from
IDL files (the Perl IDL parsing code may be a useful base
for that), and submits the necessary fixes to bring the code
in ORBit up-to-date, I'd take them, but it isn't a high priority
for me.

Regards,
                                        Owen