[gfortran] patch was: Re: Untabified and resubmitted: Patch to gfortran PR13742

[Tobias Schlüter]

Victor Leikehman wrote:
> This patch fails on the following (valid) testcase:
> 
> program main
> common /a/ x
> real :: x
> print *,x
> end
> block data
> common /a/ x
> data x /1.2345/
> end
> 

It's funny that something this simple breaks, even though I could 
compile several packages without problems related to that patch. Looks 
like code which uses common blocks tends to also use implicit typing. 
Thanks for finding this.

I found another case I hadn't thought of in my patch. I'll submit an 
updated patch once I understood how to find the common block a symbol is 
associated with. Or does anybody know? Do I have to traverse all common 
blocks in a given namespace or is there a field in gfc_symbols which I 
didn't understand?

Thanks,
- Tobi