[Patch,Fortran] PR45186 - P2 regression - gfortran 4.6 emits wrong line numbers

[Tobias Burnus]

  Dear Paul,

On 09/02/2010 03:29 PM, Paul Richard Thomas wrote:
> On Thu, Sep 2, 2010 at 3:21 PM, Tobias Burnus<burnus@net-b.de>  wrote:
>>   This is part 1 of the fix for PR fortran/45186; it changes a couple of
>> "fold_build[0-9]" to "fold_build[0-9]_loc(input_location,".
>>
>> Remaining: 622 folds_builds, namely
> Gulp!

Well, we still have a couple of weeks until 4.6.0 is released ...


> Incidentally, my manifestation of FX's memory
> checking patch, set gfc_current_locus in trans.c.  I do not have the
> source to hand at the moment.  Has this been implemented anyway?

I guess it has not; the last memcheck patch is available from 
http://gcc.gnu.org/wiki/GFortranPatchTracker

Thanks for the pointer! I think the loc part of the patch should be 
included when updating fold_build* in trans.c.

By the way, I believe we need to revisit the memcheck patch to implement 
checking for
    deallocate(ptr)
to give an error if ptr is not associated with an anonymous target (via 
allocate) but to a named target. I think that is best done via 
libgfortran and a hash. To do so is a Fortran 90/95 correctness issue, 
but it can nicely be combined with adding optional memchecking support.

Tobias