[cfe-dev] [libc++] Is namespace "index" a special kind of a namespace in libc++?

[Klemen Forstnerič via cfe-dev]

Thanks for your reply. When I try to build my previously posted example
(index.hpp and index.cpp) with the flags "-U_GNU_SOURCE -D_ISOC99_SOURCE"
it works, but when I try it on my production code, I get random errors in
all sorts of places. Having tried that, is there anything else I can do,
besides renaming my namespace index to something else?

Should I have rebuilt libc++ itself with those flags instead of my code?

I should note, though, using the command "clang++ index.cpp" (which uses
the system's libstdc++ 4.8) to build my previously posted example, doesn't
require me to undefine _GNU_SOURCE and define _ISOC99_SOURCE to make it
compile. What does libstdc++ do differently than libc++?

Thanks again.

On Tue, Oct 31, 2017 at 12:04 AM, Richard Smith <richard at metafoo.co.uk>
wrote:

> On 30 October 2017 at 07:52, Klemen Forstnerič via cfe-dev <
> cfe-dev at lists.llvm.org> wrote:
>
>> Hi Richard,
>>
>> thanks replying. So if I understood correctly, undefining _GNU_SOURCE
>> and defining _ISOC99_SOURCE instead, should be a good fix until this
>> gets patched in libc++? :-)
>>
>
> There's nothing that libc++ can do about it; the 'index' function is
> coming from your libc implementation. But yes, undefining _GNU_SOURCE and
> defining _ISOC99_SOURCE (and maybe also _ISOC11_SOURCE) should probably
> work.
>
>
>> Cheers,
>> Klemen
>>
>> On Sun, Oct 29, 2017 at 11:14 PM, Richard Smith <richard at metafoo.co.uk>
>> wrote:
>>
>>> Looks like glibc provides it in <string.h> too if _GNU_SOURCE is
>>> defined, which it *always* is when compiling in C++ mode (unless explicitly
>>> undefined). :-(
>>>
>>> Is it still the case that glibc requires _GNU_SOURCE to be defined in
>>> order for it to provide the symbols that C++ compilation requires? A quick
>>> test seems to show that _ISOC99_SOURCE (or sometimes _ISOC11_SOURCE) is
>>> actually sufficient. Perhaps we should provide a mode that doesn't define
>>> _GNU_SOURCE (ideally, -std=c++XX would have that effect, but that's
>>> probably going to break too much).
>>>
>>>
>>> On 29 October 2017 at 13:14, Richard Smith <richard at metafoo.co.uk>
>>> wrote:
>>>
>>>> There is a ::index function in strings.h on some platforms that might
>>>> be leaking into <algorithm>.
>>>>
>>>> On 29 Oct 2017 13:10, "Klemen Forstnerič via cfe-dev" <
>>>> cfe-dev at lists.llvm.org> wrote:
>>>>
>>>> Hi libc++ developers,
>>>>
>>>> I'm having a problem compiling the following code:
>>>>
>>>> Header (index.hpp):
>>>>
>>>> #ifndef INDEX_HPP_
>>>> #define INDEX_HPP_
>>>>
>>>> #include <algorithm>
>>>>
>>>> namespace index {
>>>>
>>>> class Index {
>>>>  public:
>>>>   Index();
>>>> };
>>>>
>>>> }
>>>>
>>>> #endif // INDEX_HPP_
>>>>
>>>> Source (index.cpp):
>>>>
>>>> #include "index.hpp"
>>>>
>>>> namespace index {
>>>>
>>>> Index::Index() {
>>>> }
>>>>
>>>> }
>>>>
>>>> These are the errors I get:
>>>>
>>>> index.cpp:5:1: error: use of undeclared identifier 'Index'; did you
>>>> mean '::index::Index'?
>>>> Index::Index() {
>>>> ^~~~~
>>>> ::index::Index
>>>> ./index.hpp:8:7: note: '::index::Index' declared here
>>>> class Index {
>>>>       ^
>>>> index.cpp:5:8: error: cannot define or redeclare 'Index' here because
>>>> namespace 'index' does not enclose namespace 'Index'
>>>> Index::Index() {
>>>> ~~~~~~~^
>>>> 2 errors generated.
>>>>
>>>> This is my compile command:
>>>> clang++ --stdlib=libc++ index.cpp
>>>>
>>>> clang version:
>>>> clang version 5.0.0 (tags/RELEASE_500/final)
>>>> Target: x86_64-unknown-linux-gnu
>>>> Thread model: posix
>>>> InstalledDir: /usr/local/bin
>>>>
>>>> libc++ version is today's trunk
>>>>
>>>> System: Linux trusty64 3.16.0-55-generic #74~14.04.1-Ubuntu SMP Tue Nov
>>>> 17 10:15:59 UTC 2015 x86_64 x86_64 x86_64 GNU/Linux
>>>>
>>>>
>>>> If I don't include <algorithm> or if I change the name of the namespace
>>>> to something other than "index", the code compiles.
>>>>
>>>> Is this a libc++ bug?
>>>>
>>>> Cheers,
>>>> Klemen
>>>>
>>>> _______________________________________________
>>>> cfe-dev mailing list
>>>> cfe-dev at lists.llvm.org
>>>> http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-dev
>>>>
>>>>
>>>>
>>>
>>
>> _______________________________________________
>> cfe-dev mailing list
>> cfe-dev at lists.llvm.org
>> http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-dev
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20171031/cf29495b/attachment.html>