Re: flex -- input buffer overflow problem

[Nils Nordman]

On Mon, Oct 22, 2001 at 03:19:11PM +0200, Hans-Bernhard Broeker wrote:
> 
> Only a very vague idea: default rule being matched in a scanner that uses
> several different start conditions --- this could mean that the start
> condition isn't always initialized to INITIAL reliably when a new file
> gets started. Particularly since that scanner doesn't issue a single
> forced BEGIN(some_state), but rather relies entirely on yy_push_state()
> and yy_pop_state() for start condition changes.

Spot on. It seems some startcondition does not get popped correctly.
Having a BEGIN(0) just before scanning a new file eliminates this
problem completely.

Thanks! 

-- 
Nils Nordman <address@hidden>