Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!rutgers!rochester!pt.cs.cmu.edu!cadre!pitt!cisunx!msw
From: msw@cisunx.UUCP (Matt S Wartell)
Newsgroups: comp.std.c
Subject: Re: End user implications of trigraphs (SUMMARY)
Summary: summary of responses to question about ANSI trigraphs
Keywords: trigraphs
Message-ID: <17903@cisunx.UUCP>
Date: 3 May 89 13:30:49 GMT
References: <17806@cisunx.UUCP>
Reply-To: msw@unix.cis.pittsburgh.edu (Matt S Wartell)
Followup-To: junk
Distribution: na
Organization: University of Pittsburgh
Lines: 27

Thanks to all who responded to my somewhat naive question about trigraphs.
As was pointed out by all of the respondants, my question was ill-formed.
I had asked if
	puts("trigraph??");
would break under the pANS.  Since ??" is not a ISO-646 trigraph, my example
would work fine.  If the question were:
	puts("trigraph??!");
the answer would be: yes, this will generate "trigraph|" instead of 
"trigraph??!".

It was also pointed out that ISO-646 trigraphs _have_ been incorporated
into the proposed standard.  For what it's worth, trigraphs are here.

However, a few people noted that self-respecting ANSI compilers would
probably have a "disable-trigraphs" or "enable-trigraphs" option.

A helpful solution came from Karl Heuer via Mark Brader's .signature:
	sed -e "s;??\\([-=(/)'<!>]\\);?\\\\?\\1;g"
will protect trigraphs-that-weren't-meant-to-be-trigraphs in existing
code.  Thus, the ??! from my example would become ?\?! when transformed
by sed.  One caveat: the ?\? construction may fail on pre-ANSI compilers.

Thanks again to Karl Heuer, Marv Rubinstein, Gordon L. Burditt, 
Stephen J. Friedl, Mark Brader and David H. Wolfskill for their kind
responses. 
-- 
matt wartell, university of pittsburgh             msw@unix.cis.pittsburgh.edu