Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!uunet!husc6!cmcl2!rutgers!nysernic!itsgw!batcomputer!cornell!svax!peckham
From: peckham@svax.cs.cornell.edu (Stephen Peckham)
Newsgroups: comp.unix.wizards
Subject: Lex Bugs?
Message-ID: <1729@svax.cs.cornell.edu>
Date: Wed, 21-Oct-87 11:38:56 EDT
Article-I.D.: svax.1729
Posted: Wed Oct 21 11:38:56 1987
Date-Received: Fri, 23-Oct-87 07:44:52 EDT
Reply-To: peckham@svax.cs.cornell.edu (Stephen Peckham)
Distribution: comp
Organization: Cornell Univ. CS Dept, Ithaca NY
Lines: 29

I've run across two problems using lex that appear to be bugs.
First, if the first rule in the lex specification is "\<|" (without quotes), 
the generated code won't compile. (A "case" outside of
a switch is generated.)  No lex error is given for the "|" operator
followed by the end-of-line.  Second, a specification with the lookahead
operator doesn't work correctly, and depends on the order of rules.
An excerpt follows:

%%
<W>([\ \n]*)/("<") {BEGIN C;}
<W>([\ \n]*)       {BEGIN S;}
<C>"<"   
<INITIAL>"<|"      {BEGIN W;}

On input "<| <..." the following tokens are returned:
"<|" (rule 4)
" " (rule 1)
"<" (rule 1)--this sure looks fishy.

If I move rule 3 (<C>...) to the beginning of the specification,
then the third token returned is "<", and it is returned by the
moved rule.  Of course, this rule should have matched in the first place.

I'm running a Sep 15 1986 version of lex on a Sun 3/50.
Are these known bugs, or am I misunderstanding something?
If someone is interested, I can send more details.

Steve Peckham
peckham@svax.cs.cornell.edu