[PATCH v4 6/7] Replace alloc+strcpy+strcat with asprintf() & XNFasprintf() calls

[walter harms]

Am 06.12.2010 09:34, schrieb Alan Coopersmith:
> walter harms wrote:
>>
>> Am 05.12.2010 18:51, schrieb Alan Coopersmith:
>>
>>>>> diff --git a/hw/xfree86/common/xf86ShowOpts.c b/hw/xfree86/common/xf86ShowOpts.c
>>>>> index ce86090..c0fa80a 100644
>>>>> --- a/hw/xfree86/common/xf86ShowOpts.c
>>>>> +++ b/hw/xfree86/common/xf86ShowOpts.c
>>>>> @@ -97,11 +97,8 @@ void DoShowOptions (void) {
>>>>>  				);
>>>>>  				continue;                                                       
>>>>>  			}
>>>>> -			pSymbol = malloc(
>>>>> -				strlen(xf86DriverList[i]->driverName) + strlen("ModuleData") + 1
>>>>> -			);
>>>>> -			strcpy (pSymbol, xf86DriverList[i]->driverName);
>>>>> -			strcat (pSymbol, "ModuleData");
>>>>> +			XNFasprintf(&pSymbol, "%sModuleData",
>>>>> +				    xf86DriverList[i]->driverName);
>>>>
>>>> every code before checks the return value of asprintf but not here ?
>>> We don't need to know the length, and error is not an option from the XNF*
>>> variants - they either succeed or cause the server to abort.   I used the
>>> XNF* here since if the malloc failed, previously the server exited, just
>>> going through the sigsegv handler instead of a cleaner AbortServer().
>>> (NF is "No Fail")
>>>
>>
>>
>> mmh, so XNFasprintf() could be void instead of int ?
> 
> If we were sure that no caller would ever want the returned string length - but
> we're not, and that would be a needless change from the normal asprintf() API.
> 

but this is not the normal asprintf() but you are right it makes sense to have a length
so you can use realloc to increase it, or you drop it and use
asprintf(&mod,"%s%",mod,mod2) to append.
Do you realy think anyone whould use it ? people we use it as errorindicator and what
value has an errorindicator in a function that can not fail ?

re,
 wh