Re: [fortran, patch] Document the interaction of -finit-{real,integer,…} and -Wuninitialized

[Tobias Burnus]

On 11/08/2011 08:55 AM, FX wrote:
> The attached patch documents the fact that using the -finit-* options will silence -Wuninitialized warnings.

Good idea. (And approved by Janne.)

> I don't think that's unreasonable [...], so I intend to close the PR after committing.

I think one could distinguish between front-end warnings, where the FE 
keeps track of uninitialized variables, and middle-end warnings. The 
-finit-* flags surely have to silence ME warnings. However, they should 
not affect FE warnings. I think we essentially do not have any FE 
uninitialized diagnostic, but one could consider adding one. That would 
help in some cases [like this one - or when the ME lacks information 
about the Fortran semantic], though I do not want to do a full flow 
analysis. It would be more like the current FE "variable is unused" 
warning where one simply sets a flag when the value of a variable gets 
references (for unused) or set (for uninitialized), one could even 
combine it to set but not used.

Tobias