Re: [PATCHv2 2/2] Fix sparse warnings

[Stephen Boyd <bebarino@gmail.com>]

On Mon, Mar 21, 2011 at 2:58 PM, Junio C Hamano <gitster@pobox.com> wrote:
> Junio C Hamano <gitster@pobox.com> writes:
>
>> Still yeek...
>>
>> What I meant was more like at the minimum:
>> ...
>> or much more preferably:
>>
>>  - These files use symbols without declaring, because they do not include
>>    "builtin.h":
>>
>>     builtin/clone.c (cmd_clone), builtin/fetch-pack.c (cmd_fetch_pack), ...
>>
>>  - These files define extern symbols without declaring, and they can be
>>    file scope static:
>>
>>     builtin/fmt-merge-msg.c (init_src_data), ...
>>
>>  - These callsites pass literal integer 0 where they mean to pass a NULL
>>    pointer:
>>
>>    builtin/notes.c (resolve_ref), ...
>>
>> The patch text itself look more or less Ok, but I see you have builtin.h
>> not as the first include in builtin/pack-redundant.c.
>>

Ah ok, I can do that.

>
> I spotted these two.  thread-utils.h already includes pthread.h, and
> builtin.h should come before (though technically exec_cmd.h does not
> depend on any external types, so this is just a conformity issue, not
> correctness one).
>
> Again, thanks.
>
>  builtin/pack-redundant.c |    2 +-
>  thread-utils.c           |    1 -
>  2 files changed, 1 insertions(+), 2 deletions(-)
>
> diff --git a/builtin/pack-redundant.c b/builtin/pack-redundant.c
> index 760b377..a15e366 100644
> --- a/builtin/pack-redundant.c
> +++ b/builtin/pack-redundant.c
> @@ -6,8 +6,8 @@
>  *
>  */
>
> -#include "exec_cmd.h"
>  #include "builtin.h"
> +#include "exec_cmd.h"
>
>  #define BLKSIZE 512
>
> diff --git a/thread-utils.c b/thread-utils.c
> index 2c8c1e3..7f4b76a 100644
> --- a/thread-utils.c
> +++ b/thread-utils.c
> @@ -1,5 +1,4 @@
>  #include "cache.h"
> -#include <pthread.h>
>  #include "thread-utils.h"
>
>  #if defined(hpux) || defined(__hpux) || defined(_hpux)
>

Ok, I'll squash these in and resend tonight when I get home.

Also, I don't think exec_cmd.h is actually used in some of the builtin
C files (due to some setup fallouts) so I think we can probably just
remove the exec_cmd.h includes if they're within contex and unused.
I'll do that next round.