For review.
Fixes the two ICEs reported in PR that remained after the previous fix.
There is a side affect that is manifested in the tree dumps. Instead of
__builtin_free (ptr2->dat.data);
we get
__builtin_free ((void *) ptr2->dat.data);
I do not know the cause of this but from what I can tell the newly
inserted cast is harmless. All the examples I've seen so have the cast
except where the parameter is declared as void *. In the tree dumps ptr2
is declared as struct testtype2 *, I do not know where the type is
declared so I don't know whether data is declared void * (I expect it is).
Is it worth the effort to determine how to remove the extra (void *)?
[PATCH] Fortran : Two further previously missed ICEs PR53298
There were 3 ICEs with different call stacks in the comments of this
PR. A previous commit fixed only one of those ICEs.
The ICEs fixed here are in trans-array.c and trans-expr.c.
The first ICE occurred when the array reference is not AR_ELEMENT
gfc_conv_scalarized_array_ref is called with se and ar, if se->ss is
NULL the ICE occurs. If se->ss is NULL there is nothing to do before
the return.
The second ICE occurs in code that did not match its comments. Fixing
the code to match the comments fixes the ICE. A side affect is that
the in the tree dumps for finalize_35.f90 and finalize_36.f90 contain
"__builtin_free ((void *) ptr2->dat.data);", the "(void *)" was
previously omitted. The cast is harmless.
2020-09-11 Mark Eggleston <markeggles...@gcc.gnu.org>
gcc/fortran/
PR fortran/53298
* trans-array.c (gfc_conv_array_ref): In the body of the if
statement only execute the code before the reurn is se->ss is
set.
* trans-expr.c (gfc_conv_component_ref): Change the if
expression to match the comments.
2020-09-04 Mark Eggleston <markeggles...@gcc.gnu.org>
gcc/testsuite/
PR fortran/53298
* gfortran.dg/finalize_35.f90: Handle extra (void *).
* gfortran.dg/finalize_36.f90: Handle extra (void *).
* gfortran.dg/pr53298_2.f90: New test.
* gfortran.dg/pr53298_3.f90: New test.
--
https://www.codethink.co.uk/privacy.html
