I'm checking this in as obvious.
I was looking at instances of "alloc.*sizeof" and noticed a couple
where the types in question are incorrect.
In gdbtypes, the code allocates sizeof(int) to represent a struct rank.
In mi-cmds, the code uses "struct mi_cmd **" -- one "*" too many.
In both cases the problems are latent because in practice the sizes
are the same as the sizes of the correct types. Still, it's better to
be correct.
I think gdb would be improved by a wholesale change from explicit
sizeofs to using the libiberty.h allocation macros. In most cases
they are both shorter and have better type safety. However, the
resulting patch is rather large.
Built and regtested on x86-64 Fedora 20.
2014-05-19 Tom Tromey <tromey@redhat.com>
* gdbtypes.c (rank_function): Use XNEWVEC.
* mi/mi-cmds.c (build_table): Use XCNEWVEC.
+2014-05-19 Tom Tromey <tromey@redhat.com>
+
+ * gdbtypes.c (rank_function): Use XNEWVEC.
+ * mi/mi-cmds.c (build_table): Use XCNEWVEC.
+
2014-05-19 Doug Evans <dje@google.com>
* dwarf2read.c (struct dwarf2_per_objfile): Delete unused members
bv = xmalloc (sizeof (struct badness_vector));
bv->length = nargs + 1; /* add 1 for the length-match rank. */
- bv->rank = xmalloc ((nargs + 1) * sizeof (int));
+ bv->rank = XNEWVEC (struct rank, nargs + 1);
/* First compare the lengths of the supplied lists.
If there is a mismatch, set it to a high value. */
int nr_rehash = 0;
int nr_entries = 0;
struct mi_cmd *command;
- int sizeof_table = sizeof (struct mi_cmd **) * MI_TABLE_SIZE;
- mi_table = xmalloc (sizeof_table);
- memset (mi_table, 0, sizeof_table);
+ mi_table = XCNEWVEC (struct mi_cmd *, MI_TABLE_SIZE);
for (command = commands; command->name != 0; command++)
{
struct mi_cmd **entry = lookup_table (command->name);