On 17-11-2016 01:15, Sage Weil wrote:
> On Thu, 17 Nov 2016, Willem Jan Withagen wrote:
>> On 16-11-2016 23:33, Gregory Farnum wrote:
>>> On Wed, Nov 16, 2016 at 2:29 PM, Adam C. Emerson <aemerson@xxxxxxxxxx> wrote:
>>>> On 16/11/2016, Gregory Farnum wrote:
>>>>> Is that branch going to merge imminently? If not maybe we could pull
>>>>> out this and the BlueStore fixes for fast merging. :)
>>>>
>>>> It is in wip-sage-testing at the moment.
>>>>
>>>>> Is this different from the boost changes you mentioned you already
>>>>> have? :)
>>>>
>>>> Yes and no. The boost changes fix everything in external dependencies
>>>> I know about. I was just pointing out that we shouldn't have trouble
>>>> addressing future problems if they crop up.
>>>
>>> Okay, so once your stuff is in we ought to be able to add -Werr to the
>>> build flags. Hurray!
>>
>> 'mmmm,
>>
>> Perhaps you can fix the CMakeList.txt such that ATM it is only set for
>> Linux/GCC....
>>
>> I'm all for it, and there is already quite some work done to get the
>> code to pass with CLANG.
>> But you are going to rain a bit on my parade in getting up to running
>> all tests successful and getting to a milestone for the port to FreeBSD.
>> Due to execluding some code for FreeBSD, I end up with quite a few more
>> unused errors. So I'll have to start fixing them as well.
>>
>> Perhaps it would be possible to have jenkins test compilation with Clang
>> on Linux, since otherwise my port is the only thing that is going to
>> catch Clang warnings. And it is not always easy to get those fixed if
>> you are not into the changed code.
>
> Once the clang build is working I'd love to have a clang build test!
I would not hold my breath, but we are slowly getting there.
For the code I'm compiling atm. there are remaining 3 problems for me,
and I haven't looked real hard into either one.
1) A illegal access in unittest_denc where seemingly an incorrect offset
is used. I just ignore the test result ATM
2) test/test_mempool.cc:
// {
// mempool::unittest_2::map<int,obj> v;
// v[1] = obj();
// v[2] = obj(2);
// v[3] = obj(2, 3);
// }
// }
And
// TEST(mempool, unordered_map)
// {
// mempool::unittest_2::unordered_map<int,obj> h;
// h[1] = obj();
// h[2] = obj(1);
// }
And here Clang complains about not matching types for the 2 components
during the compare function.
I just patch the code to not compile the code.
3) msg/async/AsyncMessenger.cc
in Processor::rebind
- return bind(addr, new_avoid);
+ return bind(msgr->my_inst.addr, new_avoid);
If I do not replace that reference, I end up with nonce mismatches once
connections have restarted. Doing this seemingly equivalent replacement
fixes that.
--WjW
--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
