> On Wed, 21 May 2008 17:58:16 +1000, Mark Goodwin wrote:
>
> markgw> Nathan Scott wrote:
> >> Hi,
> >>
> >> I came across an oddity in pmnscomp.c (the binary PMNS
> >> compiler) - the -v option contains conditional code on
> >> the BUILDTOOL macro, which conditional makes version 2
> >> available. This is not enabled anywhere, probably was
> >> on IRIX. Version 2 looks to be similar to version 1,
> >> except has added a checksum.
> >>
> >> Anyone know what the history is here? Why aren't we
> >> enabling version 2 everywhere?
> Wasn't it something Mr. X Moore did as his first project?
CC'd - Daniel?
> markgw> don't know.
>
> markgw> Irrespective, we should nuke compiling the pmns and just
> markgw> deal with ascii. The binary format was introduced eons ago
> markgw> for performance reasons that are no longer relevant on
> markgw> modern h/w.
> Brave assertion. Did anyone ever tried to measure how long pmcd takes
> to load pmns? I didn't.
>
> Plus, compiling is used as validation step to make sure pmns we've got
> makes any sense at all. If anything, this needs to be preserved.
I think (maybe James can confirm?) that on Mac OS X, unless the developer
tools are installed (not the default IIRC), then there's no cpp binary there.
As an initial step, we can safely remove the BUILDTOOL ifdefs, to at least
make v2 available, then as a secondary step we can consider switching to
v2 by default... it looks like libpcp has known about v2 for a very long
time,
so there should be no compatibility issues there AFAICT.
cheers.
--
Nathan
|