This is the mail archive of the
fortran@gcc.gnu.org
mailing list for the GNU Fortran project.
Re: [PATCH, Fortran] PR fortran/77584 Re: Unclassifiable statement error with procedure pointer using template named "structure_"
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=77584#c2
Committed revision 240230.
---
Fritz Reese
On Fri, Sep 16, 2016 at 2:40 PM, Steve Kargl
<sgk@troutmask.apl.washington.edu> wrote:
> On Wed, Sep 14, 2016 at 10:37:15AM -0400, Fritz Reese wrote:
>>
>> The attached patch fixes PR 77584. The match code in
>> gfc_match_decl_typespec for STRUCTURE/RECORD was a bit greedy and
>> ambiguous.
>>
>> For RECORD, the matcher was too broad in its matching and would catch
>> anything starting with 'record'. I made the code more robust here so
>> that it looks for 'record /' specifically, and keeps the helpful
>> warnings for bad record declarations if -fdec-structure is enabled.
>>
>> For STRUCTURE, the matchers were equally too broad and captured
>> anything starting with 'structure'. Unfortunately we can't easily
>> disambiguate it by capturing 'structure /%n/' because a structure
>> declaration-type-spec might be an anonymous nested structure
>> declaration, and these have no name ('/%n/'). However a
>> declaration-type-spec is only valid for structures within other
>> structures anyway (nested and anonymous structures) - otherwise we
>> would be in gfc_match_structure_decl directly through decode_statement
>> in parse.c. So instead of returning MATCH_ERROR when we see
>> 'structure' and are not nested inside a structure, we make sure we are
>> nested in a structure first and pass silently if not.
>>
>> Patched compiler passes tests on x86_64-redhat-linux. With no
>> complaints I will commit to trunk in the coming days.
>>
>> ---
>> Fritz Reese
>>
>> 2016-09-14 Fritz Reese <fritzoreese@gmail.com>
>>
>> PR fortran/77584
>>
>> * gcc/fortran/decl.c (match_record_decl, gfc_match_decl_type_spec):
>> Fixes to handling of structure/record from declaration-type-spec.
>> * gcc/testsuite/gfortran.dg/dec_structure_15.f90: New testcase.
>
> Patch looks ok to me.
>
> --
> Steve