Re: Flex vs. POSIX 1003.2-1992 repeat operator {} precedence

[Hans Aberg]

At 14:53 -0700 2002/04/27, John W. Millaway wrote:
>The standard is not our master.

If there is something wrong with the standard, then the standard should be
changed. -- But one should still be able to write programs compatible with
the standard.

>Keep in mind that this issue is being pressed by SGI employees who have been
>tasked with assuring that the 'lex' binary distributed with IRIX, etc., is in
>conformance to UNIX98  -- a requirement mandated by the U.S. government
>for all
>unix purchases. In other words, if SGI lex doesn't support the posix semantics
>for "ab{3}", then the U.S. government can conceivably refuse to purchase SGI
>equipment.  SGI plans to modify flex to behave according to the standard, then
>distribute it as 'lex' (which they are perfectly free to do.)
>
>However, while there are pressing demands at SGI to modify flex, there is no
>such urgency elsewhere.

In view of that this is a pit that most unassuming would fall into, should
not Flex issue a warning (that it can be turned off) whenever a expression
of the form ab{n} appears in the input? For example:
  Warning, file <name>.l: ab{n} interpreted as (ab){n} (resp. a(b{n})), see
    Flex manual ...

Also, would it not be better to have simply one Flex distribution, so that
"lex" can be aliased to "flex -l". What prevents such an approach?

  Hans Aberg