[cfe-commits] [PATCH] Tagging NSLog/NSLogv with the __NSString__ format instead of __printf__ format

[Ted Kremenek]

WFM.

On Jan 27, 2012, at 6:19 AM, Jean-Daniel Dupas wrote:

> Yes, I also saw this inconsistency while writing the NSLog tests, but didn't took the time to fix it.
> If everybody is OK with the attached patch, I will apply it.
> 
> Le 26 janv. 2012 à 14:19, Jordy Rose a écrit :
> 
>> Tiny dumb pet peeve: can we change "C library function" to just "library function"? NSLog isn't a C function, and it isn't an "Objective-C library function" in the same way as, say, objc_getAssociatedObject is.
>> 
>> (separate commit, since it's an unrelated change...it just caught my eye now)
>> 
>> Jordy
>> 
>> 
>> On Jan 24, 2012, at 18:03, Jean-Daniel Dupas wrote:
>> 
>>> 
>>> Le 24 janv. 2012 à 08:07, Ted Kremenek a écrit :
>>> 
>>>> On Jan 21, 2012, at 3:49 AM, Jean-Daniel Dupas <devlists at shadowlab.org> wrote:
>>>> 
>>>>> I have a first patch that add NSLog/NSLogv to Builtins.def, but instead of adding a new letter to define NSString format attribute, I enhance the AddKnownFunctionAttributes function to automatically specify "NSString" format instead of "printf" format when the format parameter is an objc object.
>>>>> 
>>>>> This patch also remove trailing '/' in front of objc header names in Builtins.def. It prevents message like "include </obj/objc-exception.h>" in diagnostics.
>>>> 
>>>> Looks good, but can we separate this into two separate commits?  They are entirely orthogonal, and a patch is easier to review when it doesn't blur together a bunch of changes for disjoint functionality into a single commit.
>>> 
>>> Of course we can. If you confirm they are fine, I will commit them.
> 
> -- Jean-Daniel
> 
> 
> 
> <clib.patch>