[swift-dev] [swift-lldb-dev] Switching swift to C++14

Ben Langmuir blangmuir at apple.com
Mon Jan 8 14:32:38 CST 2018 

Hi Ted,

I've created a PR to revert this: https://github.com/apple/swift/pull/13804 <https://github.com/apple/swift/pull/13804>

Updating to C++14 broke our Ubuntu 14.04 bots.  While those bots are using a new-enough clang, they are still using the system's libstdc++ 4.8 which doesn't include full C++14 support.  Since this ends up linked to the swift-runtime, I'm guessing it would not be practical for us to upgrade the c++ standard library for our Ubuntu 14.04 configuration since that would affect deployment of swift programs on that platform.  If that's the case, then it may be likewise impractical for us to support both Ubuntu 14.04 (pre-c++14 libstdc++) and Windows (needs c++14, IIUC).

Caveat: I am not an expert on either distributing software on Linux or Windows, I'm just making the bots happy.

Ben

> On Jan 5, 2018, at 5:54 PM, Ted Kremenek via swift-dev <swift-dev at swift.org> wrote:
> 
> Thanks Jim.
> 
> I’ve merged the PR to enable this change.
> 
>> On Jan 5, 2018, at 2:38 PM, Jim Ingham <jingham at apple.com> wrote:
>> 
>> I have no formal objections to switching lldb to C++14.  I haven't been following what is in C++14, but presumably it's purely additive?  If so then except for bugs this should be a no-op.  
>> 
>> My only reservation is it seems a little odd to adopt a "use C++14" policy for Swift lldb and not for llvm.org lldb.  Is clang going to move to C++14 some time soon?  I don't remember seeing anything about this on the lldb dev lists.  If we're going to build GitHub lldb this way it would be worth raising the suggestion on the lldb-dev as well.
>> 
>> And of course, all this is dependent on lldb actually building and passing the testsuite built with C++14.  Has somebody tried that?
>> 
>> Jim
>> 
>> 
>> 
>>> On Jan 3, 2018, at 8:46 PM, Ted Kremenek via swift-lldb-dev <swift-lldb-dev at swift.org> wrote:
>>> 
>>> I’m pinging some folks, but not everybody is back yet.  I hope to resolve this ASAP.
>>> 
>>> On Jan 2, 2018, at 11:33 AM, Saleem Abdulrasool <compnerd at compnerd.org> wrote:
>>> 
>>>> On Wed, Dec 20, 2017 at 4:26 PM, Ted Kremenek <kremenek at apple.com> wrote:
>>>> 
>>>> On Dec 19, 2017, 8:33 PM -0800, Ted Kremenek via swift-dev <swift-dev at swift.org>, wrote:
>>>>> 
>>>>> 
>>>>> On Dec 19, 2017, at 5:08 PM, Saleem Abdulrasool <compnerd at compnerd.org> wrote:
>>>>> 
>>>>>> 
>>>>>> 
>>>>>>> On Dec 19, 2017, at 3:59 PM, Ted Kremenek <kremenek at apple.com> wrote:
>>>>>>> 
>>>>>>> 
>>>>>>> 
>>>>>>>> On Dec 19, 2017, at 2:31 PM, Daniel Dunbar <daniel_dunbar at apple.com> wrote:
>>>>>>>> 
>>>>>>>> 
>>>>>>>> 
>>>>>>>>> On Dec 19, 2017, at 2:27 PM, Ted Kremenek <kremenek at apple.com> wrote:
>>>>>>>>> 
>>>>>>>>> Fair enough.
>>>>>>>>> 
>>>>>>>>> We care about swiftc and llbuild building a variety of platforms today — FreeBSD, Rasberry Pi, etc.  My impression is that C++14 is generally supported by both (a) the mininum versions of the distributions we support today and (b) the current versions of the platforms we’d like to expand Swift to in the future.  Does that sound right?  I suspect you went through the same kind of reasoning with llbuild.
>>>>>>>> 
>>>>>>>> It sounds reasonable, but to be honest I never did an audit of what platforms supported C++14.
>>>>>>> 
>>>>>>> Interesting.  Was that not a concern when that choice was made for llbuild, or was the context different?
>>>>>>> 
>>>>>>>> 
>>>>>>>> I do think that we could always use the Clang++ we build as part of Swift to build Swift itself. By that logic, it seems reasonable to expect we could always have C++14 support, although it does mean that porters would need modern Clang to support their platform. However, that is likely largely a prerequisite for Swift to work as well.
>>>>>>> 
>>>>>>> That’s a significant change to make just to get C++14 support guaranteed, and (I believe) would have a non-trivial impact on those using development environments that expect to use the Clang bundled with them to build projects.
>>>>>> 
>>>>>> Agreed.  However, the current supported platforms already support this minimum.  I think that future platforms will need to provide that anyways.  If the system doesn’t have a modern toolchain available, I suspect that it already would not have C++11 available either.  In such a scenario, they need to provide a newer toolchain, and so the difference there is minimal as we already require C++11.
>>>>> 
>>>>> OK, I’m convinced.
>>>>> 
>>>>> I want to hold off for a tiny bit just to see if anyone else has any commentary on this thread before we make a change. 
>>>> 
>>>> We’d also need to make this change on the LLDB side, and I realize now that a few people I’d like to weigh in are on vacation until the holidays.  How about we pick this up immediately in the new year once it’s clear everyone that should weigh in has had a chance to do so.
>>>> 
>>>> Just a post-holiday bump to bring this back up :-).  I'd like to get this merged so that the other Windows stuff is unblocked.
>>>> 
>>>>> 
>>>>> 
>>>>>> 
>>>>>>>> 
>>>>>>>> - Daniel
>>>>>>>> 
>>>>>>>>> 
>>>>>>>>> On Dec 19, 2017, 2:23 PM -0800, Daniel Dunbar <daniel_dunbar at apple.com>, wrote:
>>>>>>>>>> It wasn’t changed, it has *always* been C++14 since the day we open sourced it. I only investigated the platforms we officially support (Ubuntu 14.04/15.10 at the time, and macOS 10.10+ IIRC).
>>>>>>>>>> 
>>>>>>>>>> - Daniel
>>>>>>>>>> 
>>>>>>>>>>> On Dec 19, 2017, at 2:21 PM, Ted Kremenek <kremenek at apple.com> wrote:
>>>>>>>>>>> 
>>>>>>>>>>> Daniel,
>>>>>>>>>>> 
>>>>>>>>>>> When you changed llbuild to require C++14, what platforms did you take into account with that change? If you have already done the assessment here it could speed a resolution of a decision.
>>>>>>>>>>> 
>>>>>>>>>>> Thanks,
>>>>>>>>>>> Ted
>>>>>>>>>>> 
>>>>>>>>>>>> On Dec 13, 2017, at 3:19 PM, Daniel Dunbar via swift-lldb-dev <swift-lldb-dev at swift.org> wrote:
>>>>>>>>>>>> 
>>>>>>>>>>>> FWIW, llbuild requires C++14.
>>>>>>>>>>>> 
>>>>>>>>>>>> We have to do some minor shenanigans to workaround bugs in libstdc++ on 14.04, and our use is probably minimal, but just throwing that out there.
>>>>>>>>>>>> 
>>>>>>>>>>>> - Daniel
>>>>>>>>>>>> 
>>>>>>>>>>>>> On Dec 13, 2017, at 1:45 PM, Jordan Rose via swift-lldb-dev <swift-lldb-dev at swift.org> wrote:
>>>>>>>>>>>>> 
>>>>>>>>>>>>> No one else has commented on this yet today, so I'll put in that I don't have any objections to this and don't foresee any major problems. The one place where we'd need to be careful is with LLDB, which imports Swift headers; if Swift is going to move to C++14, then Swift-LLDB probably has to as well. LLDB folks, what do you think?
>>>>>>>>>>>>> 
>>>>>>>>>>>>> The other thing to check is if our minimum Clang or libstdc++ requirements on Linux didn't support C++14. It looks like our README is vague on that, but LLDB already suggests a minimum requirement of Clang 3.5, which is new enough. I suspect we're okay here.
>>>>>>>>>>>>> 
>>>>>>>>>>>>> Jordan
>>>>>>>>>>>>> 
>>>>>>>>>>>>> 
>>>>>>>>>>>>>> On Dec 13, 2017, at 10:36, Saleem Abdulrasool via swift-dev <swift-dev at swift.org> wrote:
>>>>>>>>>>>>>> 
>>>>>>>>>>>>>> Hi,
>>>>>>>>>>>>>> 
>>>>>>>>>>>>>> The newer Windows SDK requires the use of C++14 (the SDK headers use `auto` return types without trailing type information). Joe mentioned that there was some interest in switching the rest of swift to C++14 as well. I figured that I would just start a thread here to determine if this is okay to do globally rather than just specifically for the Windows builds to ensure that we can build the Windows components.
>>>>>>>>>>>>>> 
>>>>>>>>>>>>>> Thanks.
>>>>>>>>>>>>>> 
>>>>>>>>>>>>>> --
>>>>>>>>>>>>>> Saleem Abdulrasool
>>>>>>>>>>>>>> compnerd (at) compnerd (dot) org
>>>>>>>>>>>>>> _______________________________________________
>>>>>>>>>>>>>> swift-dev mailing list
>>>>>>>>>>>>>> swift-dev at swift.org
>>>>>>>>>>>>>> https://lists.swift.org/mailman/listinfo/swift-dev
>>>>>>>>>>>>> 
>>>>>>>>>>>>> _______________________________________________
>>>>>>>>>>>>> swift-lldb-dev mailing list
>>>>>>>>>>>>> swift-lldb-dev at swift.org
>>>>>>>>>>>>> https://lists.swift.org/mailman/listinfo/swift-lldb-dev
>>>>>>>>>>>> 
>>>>>>>>>>>> _______________________________________________
>>>>>>>>>>>> swift-lldb-dev mailing list
>>>>>>>>>>>> swift-lldb-dev at swift.org
>>>>>>>>>>>> https://lists.swift.org/mailman/listinfo/swift-lldb-dev
>>>>>> 
>>>>> _______________________________________________
>>>>> swift-dev mailing list
>>>>> swift-dev at swift.org
>>>>> https://lists.swift.org/mailman/listinfo/swift-dev
>>>> 
>>>> 
>>>> 
>>>> -- 
>>>> Saleem Abdulrasool
>>>> compnerd (at) compnerd (dot) org
>>> _______________________________________________
>>> swift-lldb-dev mailing list
>>> swift-lldb-dev at swift.org
>>> https://lists.swift.org/mailman/listinfo/swift-lldb-dev
>> 
> 
> _______________________________________________
> swift-dev mailing list
> swift-dev at swift.org
> https://lists.swift.org/mailman/listinfo/swift-dev

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.swift.org/pipermail/swift-dev/attachments/20180108/1b03b0cc/attachment.html>

More information about the swift-dev mailing list