Re: Mandates explicit -std=c++XY for c++ projects
On Tue, Oct 10, 2017 at 3:05 PM, Matthias Klose <doko@debian.org> wrote:
> On 10.10.2017 12:38, Mathieu Malaterre wrote:
>> Mathias,
>>
>> On Tue, Oct 10, 2017 at 12:25 PM, Matthias Klose <doko@debian.org> wrote:
>>> On 10.10.2017 11:42, Mathieu Malaterre wrote:
>>>> On Tue, Oct 10, 2017 at 11:38 AM, Matthias Klose <doko@debian.org> wrote:
>>>>> On 10.10.2017 08:45, Mathieu Malaterre wrote:
>>>>>> Dear all,
>>>>>>
>>>>>> Since the GCC 6 release [1], the default mode for C++ is now
>>>>>> -std=gnu++14 instead of -std=gnu++98. What this means is that upon
>>>>>> (re)compilation a library written for c++98 will be recompiled using a
>>>>>> different c++ standard (c++14 in this case), unless of course the
>>>>>> upstream package explicitly set the -std= flags with the appropriate
>>>>>> c++ version.
>>>>>>
>>>>>> The ISO committee generally describe the change in between different
>>>>>> standards [2] and in some case, one can find examples of subtle change
>>>>>> in behaviors [3] and [4].
>>>>>>
>>>>>> With this mind I'd like to make mandatory the -std=c++XY flags when
>>>>>> compiling either a c++ library or a stand-alone c++ program:
>>>>>>
>>>>>> 1. Either upstream define the explicit -std=c++XY flags by mean of its
>>>>>> build system,
>>>>>> 2. Or the package maintainers needs to explicit change the CXXFLAGS to
>>>>>> pass the appropriate version of the c++ standard. In which case this
>>>>>> should be documented in the README.Debian file.
>>>>>> 3. As a fallback, dh should initialize the CXXFLAGS with -std=gnu++98
>>>>>>
>>>>>> If there is a consensus on the following change, I'll go ahead and
>>>>>> also file a bug for lintian to scan the compilation logs in search for
>>>>>> missing -std=c++ expression when g++ command line are issued.
>>>>>
>>>>> I don't think this is a good idea, and I'm still trying to understand what
>>>>> problem you are trying to solve. It took a while until GCC had stable c++11
>>>>> support, and now you want to fallback to a 20 year old standard by default?
>>>>
>>>> As said above, this is simply a fallback, I am fine with any value as
>>>> long as I can see it in the log clearly.
>>>>
>>>> My point was about making the flag *explicit*, whatever value is
>>>> chosen, so that upon recompilation we gets the same symbols, the same
>>>> behavior, no FTBFS (because of deprecated feature).
>>>
>>> Various libraries do error out on deprecated functions/macros, which you can
>>> override by preprocessor macros. This is usually done in the package. Why
>>> should that be different for the compiler?
>>>
>>>>> It would be better to spend some time to prepare for -std=gnu++17 instead ;)
>>>>
>>>> So you have no major objection against my proposal,
>>>
>>> No, I have objections, because after a while this will become a debt to maintain ...
>>
>> Thanks for taking the time to answer.
>>
>> As with Gert's answer, I fail to understand your point: You would
>> prefer to see packages being (re)compiled *without* an explicit -std=
>> flag, so that it always gets recompiled using whatever is the current
>> c++ standard used by gcc, correct ?
>
> yes, that's what we do with every package, so why do you want to have an
> exception here?
True. That was the root of my confusion, a change in the c++ language
is simply just like a change in the API (or ABI).
>>>> I feared you may
>>>> have an issue with mixed combination of stdc++ runtime lib (not sure
>>>> exactly how this is handled at low level).
>>>
>>> There's only one runtime library in use. Yes, there are more or less subtle
>>> issues which we address by library package renames for issues we cannot or do
>>> not want to handle by renaming libstdc++ itself. But it's always the same
>>> runtime library used, independent of the standard.
>>
>> Ah nice.
>>
>> I may be slamming open door, but could you confirm that you see no
>> objection in maintaining a complex stdc++ package where the c++98 ABI
>> (eg. std::string) will not be used by any packages in the near future.
>
> yes, if you either make sure that everything using this library is using c++98,
> or that this library provides a layer that it can interact with both
> std::string[c++89] and std::string[c++11]. However afaik no upstream besides
> libstdc++ has done that. Again, I don't mind doing that on a package level, but
> it shouldn't be any fallback to use an old standard.
Great, this means users can compile a project using a mixture of both ABI.
All these defeat the reason for my initial proposal :)
Thanks all for your comments,
-M
Reply to: