ISO/IEC JTC1 SC22 WG21 N4250 - 2014-10-27
Jonathan Wakely, cxx@kayari.org
Teleconference information: Date: 2014-10-24 Time: 8:00am N.Am. Pacific Time Duration: 2 hours
Sutter called the meeting to order 8:10 Pacific Time.
In attendance were:
The agenda N4163 was adopted by unanimous consent.
The minutes in N4052 were adopted by unanimous consent.
Sutter included a link to previous minutes in the agenda as requested.
Sutter pointed out the business plan and convener's report is N4137. Meredith said he didn't see mention of work items for Transactional Memory or Library Fundamentals v2. Sutter explained the report was current as of August. Wong confirmed Transactional Memory was only completed more recently. Sutter pointed out that Wong is listed on the isocpp.org/std/status page and TM will be captured in the business report for next year. There is no work item for Fundamentals v2, v1 is still under ballot and the current plan is to have a follow-on TS, not a parallel one, so we need to finish the first work item before the new one starts. Even though it will have the same project number ISO now wants a new work item to be requested. Clow thought we'd already voted on a working paper. Sutter agreed but pointed out that it omits everything from v1, even though we plan to add it all back in, to avoid confusion with the document currently going through ballot.
Since Rapperswil had five teleconferences, two for proposed resolutions for core issues, two dealing with the Concepts Lite document being worked on, one for the Transactional Memory document. Good progress with both documents and both are still in process for Urbana. Miller and Sutter and Sutton had some conversations about the path for Concepts in Urbana, but CWG still have some additional wording review to do following the teleconferences. Miller expects to spend time on it in Urbana. The Transactional Memory document is a bit more stable. There are several other documents that either are on Core's plate directly or need to go through EWG once more first. The Core page in the Urbana wiki lists them, as organised by Jens. Four papers are ready for review, three more expect to see coming from EWG. In addition Core will deal with outstanding issues as time permits. As a result of not being able to alter the draft in Rapperswil there are 87 issues in "ready" state and 28 in "tentatively ready" (reflecting work done between meetings). No count of new issues since Rapperswil, an up to date list will be ready by the beginning of the meeting.
Sutter recalled that going into Rapperswil we had hoped to vote out a PDTS for Concepts and asked whether we want to try that again in Urbana. Miller agreed, and the intention would be to work on it, have core approve it, and plan to have a motion to send the Nxxxx doc with approved changes for PDTS ballot on Friday and Saturday. Miller agreed that sounds fine.
LWG will be reviewing NB comments for Library Fundamentals v1. The issues list is growing, about 130 not looked at. LWG cleared the decks in terms of papers in front of them. Expect papers coming through from LEWG but until they do will try to get through the backlog. Tried to arrange LWG telecon on Transactional Memory but noone was available.
Sutter thanked Voutilainen for agreeing to chair EWG and Stroustrup for 25 years as chair.
Voutilainen said there are 43 papers in post-Rapperswil and pre-Urbana mailing. 79 open issues. Most of the priority stuff is incoming business on T.M. There's a small handful of papers to look at before forwarding to Core. EWG needs to talk about future directions for C++17, otherwise will be business as usual. Small amount of issue handling. Very little high priority stuff in the mailing. Not under pressure to go through all of the papers. Voutilainen will create a provisional agenda next week.
Yasskin asked people to keep things moving while he was unavailable which has been done. Counting post-Rapperswil and pre-Urbana there are 55 "library" papers. Some post-Rapperswil ones have been superseded by new papers anyway. Three span multiple groups, like N4204, which involves concurrency and library together. After triage 29 papers will go straight to LEWG. 15 to LWG initially. Seven out of the post-Rapperswil mailing dropped out.
Meredith reported that Yasskin thinks some of LEWG's papers will go almost straight through to LWG. Boehm said SG1 have more, three papers for the Concurrency TS and one for the Parallelism TS that need to be looked at by LEWG. Meredith hopes to have sufficient manpower to have two parallel LEWG groups. Wong said N4180 has two small sections for LEWG as well.
Price said N4236 probably deserves hearing by EWG and LEWG, maybe Reflection SG7 first. Clow confirmed that has been sent to SG7. Voutilainen brought up N4129 on information capture which SG7 wants to send straight to LEWG, but he wants to intervene as it requires implementation support. Also LEWG is likely to be depopulated during any SG9 Ranges discussions.
Stroustrup expects discussion on contracts, as there are five related papers.
Sutter suggested that although previously evening sessions have been used mostly for informational sessions, in light of the workload and things affecting the direction for C++17, with a lot of things need design discussions, maybe we should consider official working sessions in evenings. Meredith cautiously welcomed the idea and said that recent evening sessions have benefited from there not being many so you get cross-WG participation. If we see them as cross-WG sessions that’s great, if we just extend regular WG sessions that won’t be so good.
Stroustrup said several papers to do with the future of C++ are by definition cross-WG. e.g. contracts; pattern-matching, which affects library design; a lot of concurrency stuff cuts across language and library; static assertion stuff. At some point we have to decide where dividing line is between core and library.
Action item for Sutter, Yasskin and Voutilainen to ensure there are rooms for evening sessions and to determine interest in the relevant topics.
Sutter pointed to the isocpp.org status tracking page.
We met Sept. 4-5 in Redmond. See N4199 for the minutes.
We expect to meet as much as we can throughout the meeting, as usual. I won't be able to make it Friday and Saturday, so I've asked Lawrence Crowl to take over on those days. We will need significant LEWG/LWG time.
Main agenda items (see http://wiki.edg.com/twiki/bin/view/Wg21urbana-champaign/SG1 for an evolving draft of the real agenda:)
Boehm added that he's trying to put a rough agenda on the wiki prior to the meeting. Voutilainen asked if Boehm envisions any new EWG work coming up during the meeting, or anything needing design guidance, but Boehm thinks not until the next meeting. Sutter pointed out at least one coroutine proposal is not just concerned with Concurrency & Parallelism but also affects how you write iterators and generators, so is a language evolutionary topic that EWG should be involved with. Suggested it would be useful to find out what the broader group wants before doing too much work in SG1 that takes a different direction. Boehm and Voutilainen agreed to work together on such topics rather than separately in either SG1 or EWG.
One paper to review. Expect to do that in SG2 with chair present, or EWG as fallback.
The File System DTS ballot is open until December 5th and there are no new papers on the table, so SG3 is not planning to meet in Urbana.
There are 10 "New" issues on the active issues list (N4211) which the LWG may discuss if there is time available. None appear to be high priority.
Sutter explained that work is being handled in LEWG for now. He will ask the chair to what extent LEWG still wants to use the updated asio paper, or continue using SG4 to develop things, which would need a new SG4 chair.
Voutilainen pointed out there is overlap with executors, so is related to concurrency & parallelism. Boehm needs to coordinate that discussion with Yasskin Meredith asked if we are looking to accept Kohlhoff’s new proposal as a Networking TS. Sutter's impression is we’d have LEWG design review rather than accept it right away.
Telecons every other week ongoing. Please see N4182 for minutes.
We had a good Core review of the TM wording on Sept 15. Minor changes have been incorporated in the online draft (N4179) by Jens which contains both core and library wording.
We tried having a Library review Oct 6, but nobody from Library called in (despite the announce [c++std-lib-36828] Fwd: Oct6 SG5 TM Telecon call which probably came with too short notice, or Library people are just not used to having telecon calls).
We will book a review with Library Chair early at Urbana for Tuesday to complete that review. But library folks can read N4179 (starting in Library section) to prepare now for the review.
We have a few minor updates to the TM construct design which we will pass through EWG and LEWF for which I am writing a paper for the mailing deadline. All are not expected to be controversial. This is summarized in N4180
2014: Nov Urbana meeting: LWG 2nd review (Tuesday); EWG 3rd review(Wednesday); LEWG 3rd review (Tuesday); CWG 4th review (Thursday)
2015: May Lenexa meeting:
3-6 Month PDTS Ballot: Principle Comment Stage, receive all comments
SG6 did not meet in Rapperswil. We will meet in Urbana, with continuing discussion on existing papers, and discussion on a plan for conversion paths between all types.
Work has moved from SG8 to EWG and CWG. There is some overlap with Ranges.
Stroustrup said he's keen to get language work done and move on to the library. Once we have Concepts we need a standard library with concepts. The Paolo Alto document contains much of that. Meredith said it's not clear what it means for the library, a reboot or retaining compatibility. Stroustrup replied that the current direction is not to be bug-compatible. Meredith said LWG has not considered it at all and has no time to do that. Library people should be involved early. Sutter said it's important that all the relevant people, not only LWG, are aware of discussions and can take part in them at the right time.
SG9 did not meet between meetings.
We have one paper to discuss in Urbana: N4128, Ranges for the Standard Library, Revision 1 E. Niebler, S. Parent, A. Sutton.
I expect to take half a day (time to be coordinated with LEWG) to discuss this paper in Urbana.
Clow said he needs to coordinate scheduling with Yasskin and Niebler. Sutter requested that he work with Yasskin and Voutilainen and maybe have a show of hands on Monday to gauge interest.
Expect some discussion at the meeting.
Price said there are some competing proposals that need to be discussed before approaching EWG, hopes Carruth gets a meeting set up in Urbana.
SG10 met by teleconference on August 18. The minutes can be found here:
http://www.open-std.org/pipermail/features/2014-August/000227.html
From SG10's perspective, N4200 in the pre-Urbana mailing is ready to become a new "official" revision of SD-6 on isocpp.org. Not only does it cover C++14 as it is about to be published, it also improves coverage of the features of C++11 and C++98 (specifically in that, in many implementations, support for exception handling and RTTI can be disabled).
It does *not* cover any Technical Specification; SG10 expects them to spell out their own feature-testing recommendations, since the features of a TS are optional by definition. (Naturally these recommendations should be consistent with the principles established by SG10.)
Everyone who has an interest in feature-testing is invited to review N4200 (which is red-lined relative to the published SD-6), and report any issues found at their earliest opportunity.
If any issues or objections are reported, SG10 will probably want to meet in Urbana to address them. If not, the expectation is that SD-6 will be updated following the meeting.
Sutter reported that SG11 is inactive, if there are further database papers, they will go to LEWG for the time being.
SG12 did not meet in Rapperswil, nor did we have a midterm meeting. There are 4 papers scheduled for discussion in Urbana-Champaign: N4219, N4220, N3751, N3801.
Progress since last meeting: We have one proposal. The design was reviewed by LEWG in Rapperswil and the action item was for the paper authors to produce an updated paper with initial wording to review in LEWG. He is doing so, and it will likely come in soon but it was not an on-time paper for the pre-Urbana mailing. There were no other meetings.
Expectations for this meeting: SG13 does not plan to meet separately, but the authors can be present on Friday and Saturday and would like some LEWG API review time on Friday or Saturday, if LEWG can make time – we understand the paper is late, and preference should be given to on-time papers and there are lot of those. Possibly also some initial LWG wording review if LEWG likes what it sees and LWG has some time.
Crowl wasn't in Rapperswil but discussion appears to have petered out, papers relevant to the TS didn't get reviewed. Crowl asked for a report from someone who was present.
Voutilainen said there seems to be a large amount of disagreement on the language facilities or ARBs but particularly the library wrapper, dynarray. Huge disagreement whether it's easily implementable and whether we want something like that in a TS. So in limbo regarding how to go forward.
Stroustrup pointed out there was a paper from Garcia discussing this. Voutilainen said it is being tracked as issue 137 in the EWG issues list.
Sutter explained that ballots that are closing during the meeting (Parallelism and Library Fundamentals) can be discussed so that we can start doing ballot resolution. Because ballot overlaps meeting, SC22 are allowing us to get work done during the meeting as long as we don’t produce any new paper until the ballot closes (on Saturday). SC22 secretariat will send us an unofficial copy of NB comments prior to the meeting. No updated draft will happen during the ballot period, but that ends Friday and we won’t approve any new paper until Saturday. So this can be a ballot resolution meeting. Applies to Parallelism and Library Fundamentals v1, which both conclude on Nov 7th.
Sutter reported that Marisa Peacock, the SC22 secretariat for many years, has stepped down. Sally Seitz will be the new secretariat. She used to do it and so knows the role well. Sutter has been in touch and she knows about publication of C++14 and our ballots in progress, so the change is being handle well and nothing should get lost.
SC22 had a plenary in the same week as CppCon so several chairs and the convener participated by telecon half an hour before CppCon started. They were able to mention increasing interest in C++ around the world.
Brown asked if we have a publication date for C++14. Sutter said not yet. Final text has been produced by the project editor and sent to Geneva for publication, which takes a month or two. It should be this calendar year, but we're still calling it C++14 even if ISO trips up and publishes it on January 2nd.
Miller asked if there's any progress regarding the document repository. Sutter said he had hoped to get the last mailing into LiveLink but it didn't happen yet, maybe for the post-Urbana mailing. Doing so will be the responsibility of John Spicer, as maintaining the mailings was previously Nelson's role.
No news since the last report, WG14 have not met.
Stroustrup is interested in knowing whether the concurrency work in CPLEX is likely to be compatible with C++.
Sutter said there is a paper in the pre-Urbana mailing about C and concurrency, but expects them to do something different. Their work isn't likely to go into a C standard soon, likely a TS, so there will be time for WG21 to review it. They appear to be going for language support based on Cilk and OpenMP.
Boehm said CPLEX is fairly compatible with C++ work except for syntax.
Dos Reis pointed out compatibility is very difficult, e.g. <tgmath.h> is very hard for a C++ program to process efficiently.
Stroustrup is worried that C will do something different and for the next ten years we have to be compatible with something we can't control.
Clamage said WG14 meets next week is St Louis.
Plum said a liaison report can be given in Urbana and Sutter asks Plum to be prepared to give a concise report on Monday.
Sutter referred to the isocpp.org/std/status page again which has a spreadsheet with timeline of work items. There are several ballots that might come out of Urbana. Not all will materialise, but expect to be told to open a new project for 14882 (C++17) and maybe for Library Fundamentals v2. Possibly another for Concurrency. Transactional Memory looks like the next meeting. We want Concepts to come out in Urbana. Boehm would also like a new Parallelism v2 TS with content that is disjoint to v1, the work on task regions is ready to go into a working paper. Sutter said if it's disjoint a new work item can be requested at any time, draws attention to the third tab of the spreadsheet showing progression, where asking for a new motion is row 3, now deliberately combined with the motion to create a working paper from an N-numbered paper as the initial content. Boehm said it depends on LWG but SG1 thinks SG1 is likely to have a paper ready on Friday to be approved on the Monday of the Lenexa meeting. Voutilainen can’t resist saying it seems only natural that there are multiple Technical Specifications on Parallelism happening in parallel.
Sutter asked if there is anything new to pick up from post-Rapperswil or pre-Urbana that needs special handling, that wasn’t covered already. Hard to track things given the number of papers in the recent mailings, request for a wiki centralized place where jurisdiction for a given paper can be maintained for quick reference. Sutter has a spreadshet of the papers and offers to make it shared and editable by the chairs to say which group owns each paper. Clow could update it with the triaging he's already done for library proposals. Brown requested that it be readable by others too. Voutilainen volunteered to put the information on the wiki.
Brown suggested using the opening plenary session on Monday in Urbana to process the backlog of work done at the last meeting that couldn't be voted into the draft while the ballot was underway. Sutter explained that the intentions is that on Friday and Saturday we have motions for everything we informally approved in Rapperswil (unless something comes up during the week causing them to be pulled back). Hard to do at the start of the meeting as we need a stable working paper to apply changes to and we approve that new working paper on Monday.
Sutter is to arrange meeting rooms for cross-party evening sessions. Voutilainen and Yasskin are to monitor which proposals might be of interest to wider groups and ensure discussions take place outside respective working groups.
Voutilainen to create a wiki page with the contents of the post-Rapperswil and pre-Urbana mailings and the status of each paper.
Van Winkel requested that the SG reports include the names of the study groups. Sutter suggested links to the status page.
Brown requested that 2.2.1 in the agenda be lifted out of the Liaison Reports, as the SG chairs are not liaisons. Sutter said that technically the study groups are separate and report to WG21, so it is OK.
Agenda will be in the post-meeting mailing.
The isocpp.org/std/status page has information on the next meeting, in Lenexa. Price said that if anyone has any questions they should feel free to contact him, information is in N4059.
Will be discussed at the end of the Urbana meeting, by Nelson and/or Spicer.
Sutter ended the meeting 10:06 Pacific Time.