[PATCH] Fix debuginfo in -fopenmp code (PR debug/87039)

[Richard Biener]

On November 17, 2018 4:14:58 PM GMT+01:00, Jakub Jelinek <jakub@redhat.com> wrote:
>On Sat, Nov 17, 2018 at 08:31:32AM +0100, Richard Biener wrote:
>> On November 16, 2018 10:10:00 PM GMT+01:00, Jakub Jelinek
>> <jakub@redhat.com> wrote: Can you add a comment why doing it
>differently
>> for this case is necessary or do it the same way in all cases?
>
>Do you mean in omp-expand.c or dwarf2out.c?  I believe I'm doing it the
>same

I meant in dwarf2out.c.  IIRC we have multiple calls into dwarf2out_decl and you adjust only one? 

>way in all cases (except for the grid stuff for now where I'm deferring
>to
>Martin).  In dwarf2out.c the change just makes sure that for each
>function
>its decl_function_context is handled before that function, while before
>it
>was guaranteed just for one level.
>
>>  I guess
>> there might be latent issues for nested functions inside nested
>functions?
>
>I think the change was for Ada, not really sure if one can have nested
>functions inside of nested functions.  E.g. I think in Fortran one
>can't, in
>GNU C one can.
>
>	Jakub