Hi Daniel,
Daniel Carrera:
This is a simple patch that fixes PR 4755.
It's PR 49755 - the 9 is missing in the number ...
2011-07-22 Daniel Carrera <dcarr...@gmail.com>
* trans.c (gfc__allocate_allocatable): [PR 4755] Do not fix
and the reallocate a variable that is already allocated.
Instead of "[PR 4755]" use: <tab>PR <component>/<number>. Namely:
2011-07-22 Daniel Carrera <dcarr...@gmail.com>
PR fortran/49755
* trans.c (gfc__allocate_allocatable): [PR 4755] Do not free
and the reallocate a variable that is already allocated.
Please also include the test case from the PR in the patch - with added
"! { dg-do run }" and also containing "! PR fortran/49755".
However, the bug is not fully fixed. If you try the testcase in the
bugreport, you will see that it still fails - the stat= is correctly
set, but the array shape is wrong.
The problem is that the "a.dim" setting should be done after allocation
in the inner loop - also the a.offset is misplaced. If possible, one
should also move the "D.1585" into the inner "else" block, which would
save one from using the following code in case the memory is already
allocated:
D.1585 = a.data;
...
a.data = D.1585;
The promised dump, cf. testcase in the PR.
a.dim[0].lbound = 1;
a.dim[0].ubound = 5;
a.dim[0].stride = 1;
a.dim[1].lbound = 1;
a.dim[1].ubound = 5;
a.dim[1].stride = 5;
overflow.4 = 0;
if ((logical(kind=4)) __builtin_expect ((integer(kind=8))
(overflow.4 != 0), 0))
stat.3 = 5014;
else
{
void * restrict D.1585;
if ((logical(kind=4)) __builtin_expect ((integer(kind=8))
(a.data != 0B), 0))
{
D.1585 = a.data;
stat.3 = 5014;
}
else
....
a.data = D.1585;
}
}
a.offset = -6;
if (stat.3 != 0) goto L.3;
L.3:
Tobias
