Article View: comp.lang.modula2
Article #3415Re: structured types & procedures
From: "Mark Morgan Lloyd "
Date: Sun, 24 Mar 1996 10:42
Date: Sun, 24 Mar 1996 10:42
42 lines
2562 bytes
2562 bytes
> Mark Morgan Lloyd (mark_tbu@cix.compulink.co.uk) wrote: > >> In fact FTL seems to allow OPAQUE structured types > > >This is the only compiler I've come across that does this, and it can > be >a pain if one is not prepared. Specifically (and working from > memory), it >rewrites the .sym file (derived from the .def) after > compilation of the >.mod if it finds non-pointer opaques, which can > cause intractable >dependency problems. > > On the other hand, the "make" facility finds these dependencies, > which removes the pain. There was, it is true, a bug in the > FTL "make" which sometimes forced you to do a re-make to get > it right; but this was just a bug - which could have been fixed > if the author hadn't decided that the market was too small to > justify further development - rather than a fundamental > limitation. I use MS's make, and fixed it (as far as necessary) by TOUCHing the .rel file to 1/1/80 after .def -> .sym . We attempted to get hold of the source of the Z80 compiler, but DM declined... he invited us to fund a Z80 cross-compiler (DOS-hosted) instead. The final version of the Z80 compiler isn't too bad (provided you have a ÂŁk20 logic analyser to find the nasties), although the overlay linker is severely broken (I found a work-around but almost lost my house over that one). > When I switched from FTL to TopSpeed, I found that I had to do > a great deal of code re-writing to adapt to the more > primitive (and now standard) rules about opaque types. > In my opinion this caused some elegant code to become > considerably less elegant, less readable, and more > error-prone, because of the heavier use of pointers. > I really miss the FTL-style opaque types. In my opinion, > the present restrictions on opaque types are a major > shortcoming of Modula-2. ISTR that Wirth specified that opaquely were applicable only to pointers, and that the FTL stuff documents what they are doing as an extension. I've got live code compiling under both FTL and TS, the original conversion took me about a week most of which was changing from Z80 to PC target hardware. Mark MLl [Disclaimer in resume applies]
Technical Details
Message-ID:
<Dorpq8.2Lv@cix.compulink.co.uk>Path:
arc1-ca.pugleaf.net!cix.compulink.co.uk!usenetReferences:
<4isiun$c5r@seagoon.newcastle.edu.au>Show Full Headers
Complete Article Headers
Newsgroups: comp.lang.modula2
Path: arc1-ca.pugleaf.net!cix.compulink.co.uk!usenet
From: mark_tbu@cix.compulink.co.uk ("Mark Morgan Lloyd ")
Subject: Re: structured types & procedures
Message-ID: <Dorpq8.2Lv@cix.compulink.co.uk>
Organization: KDG Mobrey Telemetry
References: <4isiun$c5r@seagoon.newcastle.edu.au>
Date: Sun, 24 Mar 1996 10:42:07 GMT
X-News-Software: Ameol