 |
 |
On Tue, 9 Dec 2003, Norman Gray wrote: > On Tue, 9 Dec 2003, Peter W. Draper wrote: > > > > Similarly, since the files par_par (for example) are always invoked as > > > PAR_PAR, is there any reason why they're not installed as such? > > > > This follows from FORTRAN itself being a case-insensive language and VMS > > having a case-insensitive file system, so include files back then were > > 'PAR_PAR' not 'par_par' (well almost, the include files were located using > > symbols, a bit like aliases, that you created using package commands like > > star_dev etc.). > > Ah right -- that makes a lot of sense. I remember symbols, and the > delights of logical names. Ah, memory.... > > What I take from that is that > > - the package commands like star_dev are in fact redundant, since > compilers have -I flags, and since one original point of them > was to give a familiar feeling to folk migrating from VMS; and > > - I should check in the _PAR, _ERR and (fio) _CMN files as upper-case, > or case-swap on installation those files which are generated with > lowercase names. > > Is that correct? More or less, the case issue could go either way (since the thing to be included is actually named in a string, so case is allowed), just as long as its consistently one or the other (to avoid the OS X issues if nothing else). Choose the one that's least work I'd say. > > Initially when we first moved to UNIX Dave Terrett produced SPT (see > > SUN/111) which editted include file statements to the full file names, but > > we found the necessity to do this a burden, so the current system was > > adopted, which does not rely on any special features of the compiler, just > > the operating system having something like a copy/link command. > > I appreciate the goal -- it depends whether we believe that -I is a > compiler special feature. If it turned out to be a problem in fact, > it would be rather easy at configure time to detect whether -I was > supported and generate the required local links if not. OK, sounds reassuring. > > > Are there any other oddnesses which have been gnawing away at folk? > > > This is the time to sort them out. > > > > Just one further point. If you go ahead and consolidate all the > > development tweaks, remember that we do have a lot of people who use these > > now and successfully develop software, even under this incredible burdon, > > it's a pain to them if you change things (for change sake). > > I appreciate this, and I will keep reminding myself of it, but I'm not > changing things for the sake of it. Putting things into the repository > and autoconfing them, we have an opportunity to `unweird' the software > set, by removing a variety of small features which are there for mostly > historical reasons. > > The aim is to make things easier for everyone, including those who > haven't started working with the software yet. Look at configure.ac > and Makefile.am in messgen -- I hope this is relatively inviting! Right I will look, but it may take a while as I'm about to go offline while they paint my room... Not expecting to be properly back (i.e. access to my PC) up until Monday. Cheers, Peter.