Re: [PATCH] Compiler fixes for ipc_mig.c, task.c, memory_object.c, vm_kern.c

[Samuel Thibault]

Shakthi Kannan, le Thu 13 Nov 2008 23:59:59 +0530, a écrit :
> --- a/kern/ipc_mig.c
> +++ b/kern/ipc_mig.c
> @@ -53,6 +53,10 @@
>  #include <device/device_types.h>
>  #include <device/ds_routines.h>
> 
> +/* Function prototypes */
> +void device_reference (device_t dev);
> +void device_deallocate (device_t dev);

Err, no, these should rather to go ds_routines.h...

> diff --git a/kern/task.c b/kern/task.c
> index 45c59c5..f788e7b 100644
> --- a/kern/task.c
> +++ b/kern/task.c
> @@ -116,7 +116,6 @@ kern_return_t task_create(
>  {
>       register task_t new_task;
>       register processor_set_t        pset;
> -     int i;
> 
>       new_task = (task_t) zalloc(task_zone);
>       if (new_task == TASK_NULL) {
> @@ -183,6 +182,7 @@ kern_return_t task_create(
>  #endif       /* MACH_PCSAMPLE */
> 
>  #if  FAST_TAS
> +     int i;
>       for (i = 0; i < TASK_FAST_TAS_NRAS; i++)  {
>           if (inherit_memory) {
>               new_task->fast_tas_base[i] = parent_task->fast_tas_base[i];

That will be rejected by non-c99 compilers. Better just #ifdef it.

> diff --git a/vm/memory_object.c b/vm/memory_object.c
> index 57dde76..002a6bb 100644
> --- a/vm/memory_object.c
> +++ b/vm/memory_object.c
> @@ -75,6 +75,9 @@ typedef     int             memory_object_lock_result_t; /* 
> moved
> from below */
>  ipc_port_t   memory_manager_default = IP_NULL;
>  decl_simple_lock_data(,memory_manager_default_lock)
> 
> +/* Function prototypes */
> +void vm_map_copy_page_discard (vm_map_copy_t copy);

Just the same: put it into vm_map.h

> index f29a014..8d85e7a 100644
> --- a/vm/vm_kern.c
> +++ b/vm/vm_kern.c
> @@ -269,7 +269,7 @@ projected_buffer_deallocate(map, start, end)
> 
>       /*Prepare for deallocation*/
>       if (entry->vme_start < start)
> -       _vm_map_clip_start(map, entry, start);
> +       _vm_map_clip_start((struct vm_map_header *) map, entry, start);

Beeerk. And actually reveals a bug: that shouldn't be just map, but
&map->hdr (which is *not* the first member of struct vm_map, there is a
lock_data_t before...), same for clip_end.

Samuel