Hi Philippe,

On 07/27/2012 10:08 AM, Philippe Nunes wrote:
> ---
>   src/smsutil.c |   37 ++++++++++++++++++++++++++++++-------
>   1 file changed, 30 insertions(+), 7 deletions(-)
>

The patch looks 'correct', but a few points below:

> diff --git a/src/smsutil.c b/src/smsutil.c
> index a541964..59eb7b1 100644
> --- a/src/smsutil.c
> +++ b/src/smsutil.c
> @@ -4090,7 +4090,7 @@ char *cbs_decode_text(GSList *cbs_list, char *iso639_lang)
>   			unsigned char unpacked[CBS_MAX_GSM_CHARS];
>   			long written;
>   			int max_chars;
> -			int i;
> +			int i, j;
>
>   			max_chars =
>   				sms_text_capacity_gsm(CBS_MAX_GSM_CHARS, taken);
> @@ -4102,12 +4102,24 @@ char *cbs_decode_text(GSList *cbs_list, char *iso639_lang)
>   			i = iso639 ? 3 : 0;
>
>   			/*
> -			 * CR is a padding character, which means we can
> -			 * safely discard everything afterwards
> +			 * CR can be used as a padding character, which means
> +			 * we can safely discard everything afterwards
>   			 */
> +
>   			for (; i<  written; i++, bufsize++) {
> -				if (unpacked[i] == '\r')
> -					break;
> +				if (unpacked[i] == '\r') {
> +					/*
> +					 * check if this is a padding character or
> +					 * if it is a wanted<CR>
> +					 */
> +					for (j=i+1; j<  written; j++)

Why are you breaking the coding style here?  Specifically item M3.

> +						if (unpacked[j] != '\r')
> +							break;
> +
> +					if (j == written)
> +						break;
> +				}
> +
>
>   				buf[bufsize] = unpacked[i];
>   			}
> @@ -4135,8 +4147,19 @@ char *cbs_decode_text(GSList *cbs_list, char *iso639_lang)
>   			}
>
>   			while (i<  max_offset) {
> -				if (ud[i] == 0x00&&  ud[i+1] == '\r')
> -					break;
> +				if (ud[i] == 0x00&&  ud[i+1] == '\r') {
> +					int j = i+2;
> +
> +					while (j<  max_offset) {
> +						if (ud[j] != 0x00 || ud[j+1] != '\r')
> +							break;
> +
> +						j += 2;
> +					}
> +
> +					if (j == max_offset)
> +						break;
> +				}

Two points here.

First of all, you're again breaking the coding style item M3.  I know 
the existing code is also in violation, but then the first patch in this 
series should be fixing that violation, and then adding the new code on 
top.  I'm not advocating that you fix all of smsutil.c, but at least fix 
up the local area you will be touching first.

Secondly, mildly nitpicking, but why are you using a 'for' loop in the 
first chunk and now a while loop here?  At the very least you can make 
the code shorter and be more consistent.

>
>   				buf[bufsize] = ud[i];
>   				buf[bufsize + 1] = ud[i+1];

I'd really prefer that patches like this one that fix bugs in the 
{sms,stk}util.c are accompanied by changes / additions to the respective 
unit test files.  Especially in this case, as I assume that you have a 
set of test data that this is failing on.

That way I can immediately reference the test data, sanity check it and 
run the unit test to verify that the code and your changes are working 
as expected.  Furthermore, any regressions are more easily caught in the 
future.

Regards,
-Denis