Re: flex question

[Atsushi Kasuya]

Yes, it failed as well on the Solaris flex 2.5.4. 
I had few lines that touches yyleng. I don't call yyless().
So I removed the code that touches yyleng, but still failed the same way.

I'm constructing a simplified version of the code (this is a linker that
reads intermediate files), and will ask futher help.

Thanks,
Atsushi

----- Begin Included Message -----

>From address@hidden Thu Oct 18 10:23 PDT 2001
X-JNPR-Received-From: outside
X-Authentication-Warning: acp3bf.physik.rwth-aachen.de: broeker owned process 
doing -bs
Date: Thu, 18 Oct 2001 19:23:15 +0200 (MET DST)
From: Hans-Bernhard Broeker <address@hidden>
X-Sender: address@hidden
To: "W. L. Estes" <address@hidden>
cc: address@hidden, address@hidden
Subject: Re: flex question
MIME-Version: 1.0
X-OriginalArrivalTime: 18 Oct 2001 17:22:40.0593 (UTC) 
FILETIME=[7D6E2C10:01C157F9]

On Thu, 18 Oct 2001, W. L. Estes wrote:

> >   fatal flex scanner internal error--end of buffer missed
[...]
> > This problem happens only on one test case, and other 75 tests passes OK.
> > gcc version I'm using on the Linux system (Red Hat 7.1) is egcs-2.91.66. 
> > The same input works fine if I use 'lex' on a Sun Solaris machine.

Hmm... has this same scanner ever been used with 'flex' instead of
classical AT&T lex?  I.e. did you try "flex" on Solaris?

If not, the problem may well be that the scanner is assuming it can rely
on undocumented implementation details of AT&T lex.  One thing to watch
out for is modification of yyleng (including calls of yyless()) in any
rule that also uses yymore().  I got bitten by that one myself when
porting a time-approved lexer source from lex to flex.

-- 
Hans-Bernhard Broeker (address@hidden)
Even if all the snow were burnt, ashes would remain.



----- End Included Message -----