[comp.compilers] non-misbehavior of "Berkeley YACC"

bart%videovax.tv.tek.com@RELAY.CS.NET (Bart Massey) (07/27/90)

I have a YACC-able grammar in front of me, of about 2000 lines, which I
wrote for a complicated assembly language with a bunch of magic
preprocessing.  When I produce a parser for this grammar using either GNU
Bison or AT&T YACC, the resulting program fails to reject some inputs
containing simple syntax errors.  I didn't even realize there was a problem
until I tried producing the parser using Corbett's "Berkeley YACC", at which
point the resulting program correctly rejected these incorrect inputs.

Please excuse the vagueness of the above, but I haven't been able to reduce
this hairy kludge of mine to a simple grammar/program which displays the
problem.  My question is: under what known circumstances, if any, do both
Bison and YACC produce parsers which fail to reject incorrect programs?
Does BYACC fix any of these?

I hesitate to even ask such a vague question, but I thought I'd give it
a shot, since the underlying issue seems pretty critical to me -- I'm trusting
this parser to distinguish syntactically correct inputs from syntactically
incorrect ones...

Any comments or ideas, or should I just go crawl back under whatever rock
I crawled out from under? :-) 

					Bart Massey
					..tektronix!videovax.tv.tek.com!bart
					..tektronix!reed.bitnet!bart
-- 
Send compilers articles to compilers@esegue.segue.boston.ma.us
{spdcc | ima | lotus| world}!esegue.  Meta-mail to compilers-request@esegue.