Re: better error msg for ambiguous (de)crescendo?

[Ian Hulin]

Hi Werner,

Werner LEMBERG wrote:
> > warning: lilypond cannot interpret a (de)crescendo.
> >          MIDI output ignoring (de)crescendo starting at line <nnnn>.
> >
> > This way the diagnostic states why Lily can't do what the user has
> > coded, what it's doing as a result, and the consequences of the
> > error condition detected.
>
> This is not precise enough IMHO, since you won't get an error if you
> don't produce MIDI output...
O.K. Werner, so what problems (from a user perspective) is this error 
condition causing the code?  If we're giving a warning, this means we 
can carry on processing the code but what is produced is very probably 
not what the intends or expects.  Does this mean the graphical output 
will have a hairpin stretching over umpteen bars if there is no midi output?

What makes the (de)crescendo as coded impossible or ambiguous?  I agree, 
we need to be as precise as possible, but we need to also give the user 
as big a steer as we can towards sorting the problem.

If this condition is caught when interpreting a new \> or \< type 
lexeme, then presumably this is because there is an outstanding \< or \> 
already in effect which hasn't been terminated by a \! or specific 
dynamic marking (\f or \mp or \ppp or whatever).

SO how about signaling

warning: lilypond cannot interpret a (de)crescendo at line <nnnn>.
          An unterminated (de)crescendo is already in force
          add "\!" or specific dynamic mark.

and if midi is being produced, resignal
          MIDI output ignoring (de)crescendo starting at line <nnnn>.

Are there any other conditions which can generate this warning?  If so 
we need a specific warning message for that condition.

Cheers,
Ian