| Submitter | Thomas Koenig |
|---|---|
| Date | Oct. 16, 2010, 4:25 p.m. |
| Message ID | <1287246331.4082.7.camel@linux-fd1f.site> |
| Download | mbox | patch |
| Permalink | /patch/68047/ |
| State | New |
| Headers | show |
Comments
On 10/16/2010 09:25 AM, Thomas Koenig wrote: > Hello world, > > the attached one-letter patch fixes the fn spec for transfer_array, > because its array descriptor argument contains a pointer that can > (and will) be followed. I am not aware of any test case that this > fixes, but it is a latent bug which may be exposed by some test case, or > maybe be more aggressive optimization later. > > 2010-10-16 Thomas Koenig<tkoenig@gcc.gnu.org> > > Regression-tested. OK for trunk? > OK, Thanks, Jerry
Hello Jerry, > On 10/16/2010 09:25 AM, Thomas Koenig wrote: > > Hello world, > > > > the attached one-letter patch fixes the fn spec for transfer_array, > > because its array descriptor argument contains a pointer that can > > (and will) be followed. I am not aware of any test case that this > > fixes, but it is a latent bug which may be exposed by some test case, or > > maybe be more aggressive optimization later. > > > > 2010-10-16 Thomas Koenig<tkoenig@gcc.gnu.org> > > > > Regression-tested. OK for trunk? > > > OK, > Committed as rev. 165561. Thanks for the review! Thomas
Patch
Index: trans-io.c =================================================================== --- trans-io.c (Revision 165559) +++ trans-io.c (Arbeitskopie) @@ -366,7 +366,7 @@ gfc_build_io_library_fndecls (void) void_type_node, 3, dt_parm_type, pvoid_type_node, gfc_int4_type_node); iocall[IOCALL_X_ARRAY] = gfc_build_library_function_decl_with_spec ( - get_identifier (PREFIX("transfer_array")), ".wW", + get_identifier (PREFIX("transfer_array")), ".ww", void_type_node, 4, dt_parm_type, pvoid_type_node, integer_type_node, gfc_charlen_type_node);