treelang fix for rs6000 / treelang not proper front end

[Joseph S. Myers]

On Tue, 10 Jun 2003, Tim Josling wrote:

> development really doesn't have that in any explicit form. For example, 
> I think all front end specific stuff should be in front end specific 
> directories do f/e developers do not have to hack the main compiler. But 
> when someone added a whole lot of front end specific stuff to 
> gengtype.c, there was no written down principle you could appeal to to 
> say you shouldn't do that.

There is the written down checklist in sourcebuild.texi for the parts of a
front end, which lists all the places outside the front end directory that
might need to refer to each front end (except gcc/Makefile.in for the
testsuite targets, which no-one has got round to fixing).  I could and did
point this out when gengtype was added (with the clear preference that it
be fixed by removing front end mentions from common code, not by adding
yet another point to the checklist) and it was fixed.

The actual principle may not be documented, though the principle of not
having dependencies on particular targets in common code is; this is a
relic of the target interface but not the front end interface having been
documented historically.  But if you've been reading the lists long enough
you can point to something in the six years of list archives to document
almost any valid principle, if you can work the search engine (or
sometimes more conveniently, download all the mbox archives and grep
them).

I suggest that you submit a patch to codingconventions.html to document
this particular principle.  (Being able to point to a message you wrote
that mentions the principle and no-one contradicts is fine; being able to
point to documentation you wrote that documents the principle is better,
and in the GCC Coding Conventions it has more influence.)

-- 
Joseph S. Myers
jsm28@cam.ac.uk