native_interpret_expr may fail on real type
#Simply fortran target fails Patch#
> The patch of pr95450 added a verification for decoding/encoding checking
> The patch escalates the failure when Hollerith constant to real conversion > Since this test case is powerpc only, I think it can be moved to gcc/testsuite/gcc.target/powerpc/ppc-fortran. Well - and in text you will put in the commit log, if applicable.)
#Simply fortran target fails update#
– And you should update the comment in the testcase as Updated I assume that it won't affect other testcases, but that needs (After doing so, at least the dg-error in your testcase needs to be This will replace "Unclassifiable statement" by "Cannot simplify Thanks for fixing 'gcc/fortran/'.īut could you also include the 'gcc/fortran/' change 9:56 ` HAO CHEN GUI 1 sibling, 1 reply 6+ messages in threadĬc: Segher Boessenkool, David, Kewen.Lin, Peter Bergner 9:16 ` HAO CHEN GUI 9:29 ` Tobias Burnus 9:12 Escalate failure when Hollerith constant to real conversion fails HAO CHEN GUI * Re: Escalate failure when Hollerith constant to real conversion fails
+ tree source = native_interpret_expr (gfc_get_real_type (kind), buffer,ĭiff -git a/gcc/testsuite/gfortran.dg/pr103628.f90 b/gcc/testsuite/gfortran.dg/pr103628.f90 native_interpret_expr (gfc_get_real_type (kind), +++ -416,11 +416,14 gfc_interpret_float (int kind, unsigned char *buffer, size_t buffer_size, */ĭiff -git a/gcc/fortran/ b/gcc/fortran/ The constant will be padded or truncated. + result->representation.length, result->value.real)) + (unsigned char *) result->representation.string, result->representation.length, result->value.real) gfc_interpret_float (kind, (unsigned char *) result->representation.string, Result = gfc_get_constant_expr (BT_REAL, kind, &src->where) * fortran/ (gfc_hollerith2real): Return NULL whenĭiff -git a/gcc/fortran/ b/gcc/fortran/
* fortran/ (gfc_interpret_float): Return FAIL when Test and fixes the format problem according to Tobias's advice. The new testĬompared to last version, this version moves the mpfr_init after NULL tree It doesn't have a unique memory presentationįor some real values. The failure so that an ICE is reported when the verification fails. native_interpret_expr may fail on real typeĬonversion and returns a NULL tree then. The patch of pr95450 added a verification for decoding/encoding checking The patch escalates the failure when Hollerith constant to real conversionįails in native_interpret_expr. 9:29 ` Tobias Burnus 0 siblings, 2 replies 6+ messages in threadĬc: Segher Boessenkool, David, Kewen.Lin, Peter Bergner, Tobias Burnus */ + for (remap = expr1->ref remap remap = remap->next) + if (!remap->next & remap->type = REF_ARRAY + & remap->u.ar.Escalate failure when Hollerith constant to real conversion fails public inbox for help / color / mirror / Atom feed * Escalate failure when Hollerith constant to real conversion fails 9:12 HAO CHEN GUI In this case, + set it to AR_FULL so that gfc_conv_expr_descriptor does + not see it and process the bounds remapping afterwards explicitely. Find the last reference on the LHS and if it is an + array section ref, we're dealing with bounds remapping. gcc/fortran/trans-expr.c (revision 163310) +++ gcc/fortran/trans-expr.c (working copy) -4762,21 +4762,46 gfc_trans_pointer_assignment (gfc_expr * Sender: mailing list gcc/fortran/trans-expr.c Helo=) by with esmtpa (EximĤ.71) (envelope-from ) id 1Om2mV-0008Ub-SM X-SWARE-Spam-Status: No, hits=-1.2 required=5.0 tests=AWL, BAYES_50, X-Original-To: from ( )īy (Postfix) with SMTP id 71D64B70DD
0 kommentar(er)
