Re: separate compilation (was: A lesson for compiler warning writers)

sommerfeld@apollo.hp.com (Bill Sommerfeld)
Fri, 19 Jun 1992 17:48:31 GMT

          From comp.compilers

Related articles
Re: A lesson for compiler warning writers xjam@cork.CS.Berkeley.EDU (1992-06-18)
Re: separate compilation (was: A lesson for compiler warning writers) sommerfeld@apollo.hp.com (1992-06-19)
Re: separate compilation (was: A lesson for compiler warning writers) waldin@lcs.mit.edu (1992-06-22)
| List of all articles for this month |

Newsgroups: comp.compilers
From: sommerfeld@apollo.hp.com (Bill Sommerfeld)
Keywords: linker, design
Organization: Hewlett Packard
References: <19920609091040SEB1525@MVS.draper.com> 92-06-086
Date: Fri, 19 Jun 1992 17:48:31 GMT

xjam@cork.CS.Berkeley.EDU (The Crossjammer) writes:


      ... it is quite easy to enforce [separate compilation] consistency if you
      have a notion of an interface database (as John suggested). I don't
      remember how industrial strength it was, but CLU had exactly such a notion
      and the compiler would vigorously complain if interfaces in different
      modules did not match.


The CLU compiler I used while a student at MIT only maintained the
"interface database" for the lifetime of the compiler process. (The
compiler was an interactive program.. you fire it up, then typed in a
bunch of "compile x.clu", "compile y.clu", "compile z.clu" commands).


This broke down pretty badly if you tried to automate recompiles using
make, as it "forgot" all the cross-module state and assumed that
references to externals were all type-correct (much like C).
--


Post a followup to this message

Return to the comp.compilers page.
Search the comp.compilers archives again.