This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: mt_allocator.cc assumes sizeof(size_t) == sizeof(void *)
- From: Joel Sherrill <joel dot sherrill at oarcorp dot com>
- To: Jonathan Wakely <jwakely dot gcc at gmail dot com>
- Cc: Paolo Carlini <paolo dot carlini at oracle dot com>,GCC Mailing List <gcc at gcc dot gnu dot org>,DJ Delorie <dj at redhat dot com>,libstdc++ <libstdc++ at gcc dot gnu dot org>
- Date: Sat, 8 Nov 2014 09:03:04 -0600
- Subject: Re: mt_allocator.cc assumes sizeof(size_t) == sizeof(void *)
- Authentication-results: sourceware.org; auth=none
- References: <545CE038 dot 5000403 at oarcorp dot com> <545CE454 dot 1000201 at oracle dot com> <545CF9D3 dot 4090404 at oarcorp dot com> <CAH6eHdSPH73OKc7tuDN+tVWS9vkEqmyFVH-NbOy-ZfOnHPn7=A at mail dot gmail dot com>
On November 8, 2014 9:00:02 AM CST, Jonathan Wakely <jwakely.gcc@gmail.com> wrote:
>On 7 November 2014 16:56, Joel Sherrill wrote:
>>
>> On 11/7/2014 9:25 AM, Paolo Carlini wrote:
>>> Hi,
>>>
>>> On 11/07/2014 04:07 PM, Joel Sherrill wrote:
>>>> Hi
>>>>
>>>> On m32c-rtems, we have a build error in C++ because size_t
>>>> is 16-bits and pointers are 24 bits. m32c-elf probably does not
>>>> enable __GTHREAD support like rtems does. Since this is code
>>>> shared across targets, what is the best way to fix this?
>>> I don't know the exact opinion of the other library maintainers, but
>>> personally I consider mt_allocator an old experiment, which,
>definitely,
>>> to make any sense today would need profound changes. In particular,
>I
>>> don't think we can hope to get something useful from it when size_t
>is
>>> 16 bits and, more importantly, pointers are 24 bits. Thus, my
>>> recommendation would be just arranging for its code not to break
>>> bootstrap, nothing more than that.
>
>Agreed.
>
>> OK. I am building now with the size_t changed to uintptr_t. I tried
>to get
>> the type definition by including <cstdint> but it didn't compile.
>
>Did you try std::uintptr_t?
Yes. The only surprise was needing to include stdint.h instead of cstdint to get uintptr_t.
It built m32c-rtems and then I left a native Linux bootstrap and make check running. I will post the test results and patch Monday. Hopefully that is an OK solution.
--joel