Re: [gfortran] PATCH: Fix PR 15510

[Tobias Schlüter <tobias dot schlueter at physik dot uni-muenchen dot de>]

Paul Brook wrote:
> On Wednesday 09 June 2004 19:09, Tobias Schlüter wrote:
>
> > I'm wondering if this code shouldn't be moved to the resolution stage,
> > as this would allow for better error printing because we could print out
> > the loci where the variable is declared. Possibly we can only do this
> > after PR 15481 is fixed.
>
>
> Maybe. IIRC the C frontends currently use information from the optimizers to 
> generate unused uninitialized variable warnings. There's some discussion on 
> whether this is a good idea or not (it gives better diagnostics, but depends 
> on optimization levels). We should do the same if/when a conclusion is 
> reached.
The "variable unused" warning is optimization independent, it means the 
variable doesn't appear in an expression, so it can be front-end only. 
For the "var is never assigned a value" kind of warnings the more 
elaborate mechanism of the back-end are necessary, I agree.

- Tobi