https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92178
kargl at gcc dot gnu.org changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |kargl at gcc dot gnu.org
--- Comment #13 from kargl at gcc dot gnu.org ---
(In reply to anlauf from comment #12)
> Created attachment 55333 [details]
> Updated version of Steve's patch
>
> This patch adds evaluation of arguments that are neither constant, variables,
> or pointers from functions that may appear in a variable definition context.
>
> This appears to fix all testcases here and "regresses" only for
> gfortran.dg/assumed_type_17.f90, where the scan pattern does no longer match.
> The tree-dump shows an additional intermediate variable for the call with
> explicit parens as compared to without the patch, but the resulting assembler
> is unchanged.
>
> I was expecting trouble with finalization, but all current testcases pass.
Harald,
Thanks for finding this! I have no pr92178.diff file on any of systems, and I
quite frankly don't know if I can reconstruct the analysis I gave earlier. :-(
It's a good sign that the patch doesn't interfere with finalization.
Hopefully, Mikael or Paul can take a quick peek at the patch. I doubt
I'm allowed to approve your new patch.
