[gfortran, fix PR 25097] Additional constraints on FORALL headers

[Tobias Schlüter]

Sigh, the corrct PR number is PR 25076.

- Tobi

Tobias Schlüter wrote:
> Tobias Schlüter wrote:
>> Tobias Schlüter wrote:
>>> we hadn't yet implemented the following constraint on FORALL headers:
>>> "A subscript or stride in a forall-triplet-spec shall not contain a 
>>> reference to any index-name in the forall-triplet-spec-list in which 
>>> it appears."  This patch fixes this, using the pre-existing 
>>> gfc_find_forall_index for the purpose.  This meant moving it to the 
>>> front of the file, and calling it in a trivial fashion (which is 
>>> wasteful, but I can't imagine this ever becoming a performance 
>>> bottleneck).  I promise to remove the "gfc_" prefix, I only realized 
>>> its presence after I had cut the diff.
>>
>> Please don't waste your time on this patch yet: while removing the 
>> gfc_* prefix, I found that there already exists code which claims to 
>> address the same issue.  I'll see what it does, and why it doesn't do 
>> as advertized and report back.
> 
> Ok, here's an updated version.  It removes the old code and augments my 
> new code to also check all triplet specifications, not only later ones 
> (i.e. for (iter; ... ) for (iter = iter; ...) instead of for (iter; ... 
> ) for (iter = iter->next; ...)).  Throwing this at the testsuite 
> revealed the necessity to deal with NULL expressions in 
> find_forall_index(), which was done.
> 
> Built and testing on i386-darwin.  Ok, provided the testsuite passes?
> 
> Cheers,
> - Tobi
>