| Message ID | CAKwh3qj_XvUzB_4LQ4d3dAx11gCGMd6O9bS0-c_t7YL9EQi_YA@mail.gmail.com |
|---|---|
| State | New |
| Headers | show |
Dear Janus, Well done for sorting that out. OK for trunk. Thanks for the patch. Paul On 11 January 2015 at 14:38, Janus Weil <janus@gcc.gnu.org> wrote: > Hi all, > > this patch fixes a wrong-code regression related to operators, by > making sure that we look for typebound operators first, before looking > for non-typebound ones. (Note: Each typebound operator is also added > to the list of non-typebound ones, for reasons of diagnostics.) > > Regtested on x86_64-unknown-linux-gnu. Ok for trunk? 4.9/4.8? > > Cheers, > Janus > > > > 2015-01-11 Janus Weil <janus@gcc.gnu.org> > > PR fortran/63733 > * interface.c (gfc_extend_expr): Look for type-bound operators before > non-typebound ones. > > 2015-01-11 Janus Weil <janus@gcc.gnu.org> > > PR fortran/63733 > * gfortran.dg/typebound_operator_20.f90: New.
> Well done for sorting that out. OK for trunk. Thanks, Paul. Committed as r219440. What about the branches? Cheers, Janus > On 11 January 2015 at 14:38, Janus Weil <janus@gcc.gnu.org> wrote: >> Hi all, >> >> this patch fixes a wrong-code regression related to operators, by >> making sure that we look for typebound operators first, before looking >> for non-typebound ones. (Note: Each typebound operator is also added >> to the list of non-typebound ones, for reasons of diagnostics.) >> >> Regtested on x86_64-unknown-linux-gnu. Ok for trunk? 4.9/4.8? >> >> Cheers, >> Janus >> >> >> >> 2015-01-11 Janus Weil <janus@gcc.gnu.org> >> >> PR fortran/63733 >> * interface.c (gfc_extend_expr): Look for type-bound operators before >> non-typebound ones. >> >> 2015-01-11 Janus Weil <janus@gcc.gnu.org> >> >> PR fortran/63733 >> * gfortran.dg/typebound_operator_20.f90: New. > > > > -- > Outside of a dog, a book is a man's best friend. Inside of a dog it's > too dark to read. > > Groucho Marx
Dear Janus, Since it is a regression, by all means update the branches. We usually, propose delaying a bit but I am not convinced that this is effective for this kind of bug fix - usually, further problems take a long time to emerge. Thus, I would recommend that you get on with it. Thanks Paul On 11 January 2015 at 23:01, Janus Weil <janus@gcc.gnu.org> wrote: >> Well done for sorting that out. OK for trunk. > > Thanks, Paul. Committed as r219440. > > What about the branches? > > Cheers, > Janus > > > >> On 11 January 2015 at 14:38, Janus Weil <janus@gcc.gnu.org> wrote: >>> Hi all, >>> >>> this patch fixes a wrong-code regression related to operators, by >>> making sure that we look for typebound operators first, before looking >>> for non-typebound ones. (Note: Each typebound operator is also added >>> to the list of non-typebound ones, for reasons of diagnostics.) >>> >>> Regtested on x86_64-unknown-linux-gnu. Ok for trunk? 4.9/4.8? >>> >>> Cheers, >>> Janus >>> >>> >>> >>> 2015-01-11 Janus Weil <janus@gcc.gnu.org> >>> >>> PR fortran/63733 >>> * interface.c (gfc_extend_expr): Look for type-bound operators before >>> non-typebound ones. >>> >>> 2015-01-11 Janus Weil <janus@gcc.gnu.org> >>> >>> PR fortran/63733 >>> * gfortran.dg/typebound_operator_20.f90: New. >> >> >> >> -- >> Outside of a dog, a book is a man's best friend. Inside of a dog it's >> too dark to read. >> >> Groucho Marx
Good, I fully agree. Fortunately the patch applies cleanly to the 4.9 branch and regtests without errors. Thus I have applied it as r219475. Will do 4.8 soon. Cheers, Janus 2015-01-12 9:30 GMT+01:00 Paul Richard Thomas <paul.richard.thomas@gmail.com>: > Dear Janus, > > Since it is a regression, by all means update the branches. We > usually, propose delaying a bit but I am not convinced that this is > effective for this kind of bug fix - usually, further problems take a > long time to emerge. Thus, I would recommend that you get on with it. > > Thanks > > Paul > > On 11 January 2015 at 23:01, Janus Weil <janus@gcc.gnu.org> wrote: >>> Well done for sorting that out. OK for trunk. >> >> Thanks, Paul. Committed as r219440. >> >> What about the branches? >> >> Cheers, >> Janus >> >> >> >>> On 11 January 2015 at 14:38, Janus Weil <janus@gcc.gnu.org> wrote: >>>> Hi all, >>>> >>>> this patch fixes a wrong-code regression related to operators, by >>>> making sure that we look for typebound operators first, before looking >>>> for non-typebound ones. (Note: Each typebound operator is also added >>>> to the list of non-typebound ones, for reasons of diagnostics.) >>>> >>>> Regtested on x86_64-unknown-linux-gnu. Ok for trunk? 4.9/4.8? >>>> >>>> Cheers, >>>> Janus >>>> >>>> >>>> >>>> 2015-01-11 Janus Weil <janus@gcc.gnu.org> >>>> >>>> PR fortran/63733 >>>> * interface.c (gfc_extend_expr): Look for type-bound operators before >>>> non-typebound ones. >>>> >>>> 2015-01-11 Janus Weil <janus@gcc.gnu.org> >>>> >>>> PR fortran/63733 >>>> * gfortran.dg/typebound_operator_20.f90: New. >>> >>> >>> >>> -- >>> Outside of a dog, a book is a man's best friend. Inside of a dog it's >>> too dark to read. >>> >>> Groucho Marx > > > > -- > Outside of a dog, a book is a man's best friend. Inside of a dog it's > too dark to read. > > Groucho Marx
Index: gcc/fortran/interface.c =================================================================== --- gcc/fortran/interface.c (Revision 219429) +++ gcc/fortran/interface.c (Arbeitskopie) @@ -3706,6 +3706,8 @@ gfc_extend_expr (gfc_expr *e) gfc_user_op *uop; gfc_intrinsic_op i; const char *gname; + gfc_typebound_proc* tbo; + gfc_expr* tb_base; sym = NULL; @@ -3722,8 +3724,50 @@ gfc_extend_expr (gfc_expr *e) i = fold_unary_intrinsic (e->value.op.op); + /* See if we find a matching type-bound operator. */ if (i == INTRINSIC_USER) + tbo = matching_typebound_op (&tb_base, actual, + i, e->value.op.uop->name, &gname); + else + switch (i) + { +#define CHECK_OS_COMPARISON(comp) \ + case INTRINSIC_##comp: \ + case INTRINSIC_##comp##_OS: \ + tbo = matching_typebound_op (&tb_base, actual, \ + INTRINSIC_##comp, NULL, &gname); \ + if (!tbo) \ + tbo = matching_typebound_op (&tb_base, actual, \ + INTRINSIC_##comp##_OS, NULL, &gname); \ + break; + CHECK_OS_COMPARISON(EQ) + CHECK_OS_COMPARISON(NE) + CHECK_OS_COMPARISON(GT) + CHECK_OS_COMPARISON(GE) + CHECK_OS_COMPARISON(LT) + CHECK_OS_COMPARISON(LE) +#undef CHECK_OS_COMPARISON + + default: + tbo = matching_typebound_op (&tb_base, actual, i, NULL, &gname); + break; + } + + /* If there is a matching typebound-operator, replace the expression with + a call to it and succeed. */ + if (tbo) { + gcc_assert (tb_base); + build_compcall_for_operator (e, actual, tb_base, tbo, gname); + + if (!gfc_resolve_expr (e)) + return MATCH_ERROR; + else + return MATCH_YES; + } + + if (i == INTRINSIC_USER) + { for (ns = gfc_current_ns; ns; ns = ns->parent) { uop = gfc_find_uop (e->value.op.uop->name, ns); @@ -3772,58 +3816,9 @@ gfc_extend_expr (gfc_expr *e) if (sym == NULL) { - gfc_typebound_proc* tbo; - gfc_expr* tb_base; - - /* See if we find a matching type-bound operator. */ - if (i == INTRINSIC_USER) - tbo = matching_typebound_op (&tb_base, actual, - i, e->value.op.uop->name, &gname); - else - switch (i) - { -#define CHECK_OS_COMPARISON(comp) \ - case INTRINSIC_##comp: \ - case INTRINSIC_##comp##_OS: \ - tbo = matching_typebound_op (&tb_base, actual, \ - INTRINSIC_##comp, NULL, &gname); \ - if (!tbo) \ - tbo = matching_typebound_op (&tb_base, actual, \ - INTRINSIC_##comp##_OS, NULL, &gname); \ - break; - CHECK_OS_COMPARISON(EQ) - CHECK_OS_COMPARISON(NE) - CHECK_OS_COMPARISON(GT) - CHECK_OS_COMPARISON(GE) - CHECK_OS_COMPARISON(LT) - CHECK_OS_COMPARISON(LE) -#undef CHECK_OS_COMPARISON - - default: - tbo = matching_typebound_op (&tb_base, actual, i, NULL, &gname); - break; - } - - /* If there is a matching typebound-operator, replace the expression with - a call to it and succeed. */ - if (tbo) - { - bool result; - - gcc_assert (tb_base); - build_compcall_for_operator (e, actual, tb_base, tbo, gname); - - result = gfc_resolve_expr (e); - if (!result) - return MATCH_ERROR; - - return MATCH_YES; - } - /* Don't use gfc_free_actual_arglist(). */ free (actual->next); free (actual); - return MATCH_NO; }
Hi all, this patch fixes a wrong-code regression related to operators, by making sure that we look for typebound operators first, before looking for non-typebound ones. (Note: Each typebound operator is also added to the list of non-typebound ones, for reasons of diagnostics.) Regtested on x86_64-unknown-linux-gnu. Ok for trunk? 4.9/4.8? Cheers, Janus 2015-01-11 Janus Weil <janus@gcc.gnu.org> PR fortran/63733 * interface.c (gfc_extend_expr): Look for type-bound operators before non-typebound ones. 2015-01-11 Janus Weil <janus@gcc.gnu.org> PR fortran/63733 * gfortran.dg/typebound_operator_20.f90: New.