Define strnlen when it is not in libc

[Duncan P. N. Exon Smith via llvm-commits]

> On 2016-Nov-28, at 12:23, Chi-Hua Chen <chc.w1005+clang1404 at gmail.com> wrote:
> 
> Duncan,
> 
> On Mon, Nov 28, 2016 at 1:56 PM, Duncan P. N. Exon Smith
> <dexonsmith at apple.com> wrote:
>>> On 2016-Nov-28, at 10:55, Duncan P. N. Exon Smith <dexonsmith at apple.com> wrote:
>>>> On 2016-Nov-24, at 11:10, Joerg Sonnenberger via llvm-commits <llvm-commits at lists.llvm.org> wrote:
>>>> 
>>>> On Thu, Nov 24, 2016 at 02:06:55PM -0500, Chi-Hua Chen via llvm-commits wrote:
>>>>> Some older platforms, e.g. Mac 10.6.8, do not have strnlen. Define one
>>>>> in Support when necessary.
>>>> 
>>>> Personally, I'd prefer to just avoid it in first place. It is not really
>>>> much simpler than just using memchr.
>>> 
>>> `StringRef(...).find('\0')` seems cleaner.
>> 
>> But I'm fine with the original patch as well, since it avoids people introducing more uses of strnlen in the future...
> 
> I don't think either patch (one defines strnlen, the other using
> memchr -- I haven't submitted this one yet) would prevent people from
> introducing more uses of strnlen.

The one defining strnlen would prevent people from introducing uses of strnlen that fail on 10.6.

> 
> I have a question about the change in HeaderMap.cpp that introduced
> the first though: why is '&& Data[Len - 1]' necessary?
> https://github.com/llvm-mirror/clang/commit/839cd13b65d802f06276ff88d234419c38a44199#diff-860876c157b9f2ff75bce0d006cadc63

Looks like it's unnecessary.  Do the tests pass without it?

> 
>> assuming we actually support building on 10.6?  (There weren't any buildbot failures...)
> 
> I don't see any OS requirements for building llvm/clang. We did have
> to install a modified version of gcc on our old Mac.
> 
>> 
>>> 
>>>> 
>>>> Joerg
>>>> _______________________________________________
>>>> llvm-commits mailing list
>>>> llvm-commits at lists.llvm.org
>>>> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits
>>> 
>>