Move specification of deprecated headers to Annex D #1054

tkoeppe · 2016-11-16T15:21:21Z

No description provided.

tkoeppe · 2016-11-16T15:21:32Z

@AlisdairM, what do you think of this change?

AlisdairM · 2016-11-16T16:36:42Z

source/lib-intro.tex

-\indextext{library!C standard}%
-additional headers, as shown in Table~\ref{tab:cpp.c.headers}.%
+The facilities of the C standard library\indextext{library!C standard}
+are provided in 22 additional headers, as shown in Table~\ref{tab:cpp.c.headers}.%


I thought we were trying to avoid keeping accurate counts, as they are too much effort to maintain? Prefer just 'additional headers' without the 22?

I don't think we have a blanket rule for this. In the present case, it was easy enough to update the number, and we've previously updated (but kept) counts of things.

Happy to go with whatever you decide, just calling it out so that we make a deliberate decision, rather than accidentally fall into the old pattern by inertia.

AlisdairM · 2016-11-16T16:39:27Z

source/lib-intro.tex

@@ -1295,7 +1284,7 @@
 \ref{support.rtti}  & Type identification          &   \tcode{<typeinfo>}      \\ \rowsep
 \ref{support.exception} & Exception handling      &   \tcode{<exception>}     \\ \rowsep
 \ref{support.initlist}  & Initializer lists & \tcode{<initializer_list>} \\ \rowsep
-\ref{support.runtime} & Other runtime support     &   \tcode{<cstdalign>} \tcode{<cstdarg>} \tcode{<cstdbool>}       \\ \rowsep


I'm not sure we can editorially remove these headers from a free-standing implementation - that should require LWG signoff (which I expect should be non-contentious, but I am frequently surprised)

Good point. We can perhaps keep the headers in the free-standing implementation by adding a suitable sentence to Annex D?

If everything else is now in Annex D, I would suggest the "add this to freestanding" text belongs in Annex D as well.

I "fixed" this problem by reverting the changes to Clause 17. I noticed that we already list <strstream> as part of the library, even though we only ever mention it again in Annex D. We can do the same with these four headers.

AlisdairM · 2016-11-16T16:43:43Z

source/future.tex

@@ -134,6 +135,71 @@
 the namespace \tcode{std}.
 \end{example}

+\rSec2[ccomplex.syn]{Header \tcode{<ccomplex>} synopsis}


The moved sections should be of the form [depr.stable.name], so [depr.ccomplex.syn] here, and similar changes below.

AlisdairM · 2016-11-16T16:45:44Z

source/future.tex

+\indexlibrary{\idxhdr{stdalign.h}}%
+The contents of the header \tcode{<cstdalign>} are the same as the C
+standard library header \tcode{<stdalign.h>}, with the following changes:
+The header \tcode{<cstdalign>} and the header \tcode{<stdalign.h>} shall not


My biggest concern is providing the specification for the C header in the deprecated C++ header section, and similarly for stdbool.h. It is also deprecated to use any of the C headers by name in C++, but it still makes me more nervous about this change.

Then we have a conundrum: both headers <cstdbool> and <stdbool.h> are deprecated, yet you want the specification to be in the main text? So the specification is not deprecated, only the use is?

As I said, this was the one migration from the main text that gave me most concern - even though both uses are deprecated for different reasons. I' would be for this patch to proceed unless we have a better idea though.

I have a high-level answer to the concern: If the C headers are "deprecated" in the sense of "do not use this", then in the context of compatibility that can only be understood to mean "don't use C". And with that reading, I find it acceptable to define deprecated headers in Annex D, even from scratch.

Note that we do the same for <strstream>.

AlisdairM · 2016-11-18T02:48:14Z

source/lib-intro.tex

+\ref{support.exception}  & Exception handling        & \tcode{<exception>}        \\ \rowsep
+\ref{support.initlist}   & Initializer lists         & \tcode{<initializer_list>} \\ \rowsep
+\ref{support.runtime}    & Other runtime support     & \tcode{<cstdarg>}          \\ \rowsep
+\ref{depr.cstdalign.syn}, \ref{depr.cstdbool.syn} & Deprecated headers & \tcode{<cstdalign>} \tcode{<cstdbool>} \\ \rowsep


This row should probably drop to last, to preserve section-ordering in the table.
Other than this, I have no other reason to not land. Nice cleanup.

Good idea, done. Thanks!

jensmaurer · 2016-12-01T23:47:37Z

source/future.tex

@@ -51,6 +51,15 @@
 \rSec1[depr.c.headers]{C standard library headers}

 \pnum
+For compatibility with the C standard library\indextext{library!C standard},
+the \Cpp standard library provides headers
+\tcode{<ccomplex>}~(\ref{depr.ccomplex.syn}),


We're still in a section "C standard library headers" here, yet we keep talking about C++ headers.
Either rename this section to "Standard library headers" or add a section "C++ standard library headers" that contains the C++-only content. (I'm unsure what remains in the existing "C headers" section in the latter case.)

@jensmaurer: OK, snark alert: If you squint at it right, you could think of "C standard library headers" to be those headers in the standard library that start with "C". Then it would kind of make sense.

Ultimately, everything we describe in this document is about C++. The name "C header" is just a mnemonic. (We cannot and do not specify the C language here.)

@jensmaurer: How about "Compatibility headers"?

@tkoeppe: The entire clause talks about compatibility features.
Maybe just say "Headers"? And <ccomplex> and friends aren't really for compatibility with C, despite your edit saying so.

Hm. "Headers" sounds oddly bland, but maybe that's the only accurate description of this part...

jensmaurer · 2016-12-01T23:50:55Z

The general approach of moving the definitions of the deprecated headers into annex D is good. Annex C is about incompatibilities with C, so we can be non-normative and terse there, with plenty of \refs.

tkoeppe

See also #1054 for a similar issue.

tkoeppe · 2016-12-01T23:53:36Z

source/future.tex

@@ -51,6 +51,15 @@
 \rSec1[depr.c.headers]{C standard library headers}

 \pnum
+For compatibility with the C standard library\indextext{library!C standard},
+the \Cpp standard library provides headers
+\tcode{<ccomplex>}~(\ref{depr.ccomplex.syn}),


@jensmaurer: How about "Compatibility headers"?

jensmaurer · 2016-12-12T22:28:30Z

You need to rebase.

jensmaurer · 2016-12-15T11:35:07Z

This interacts with LWG 2835 "LWG 2536 seems to misspecify <tgmath.h>".

tkoeppe · 2016-12-15T11:36:15Z

Yes, absolutely.

tkoeppe · 2017-02-04T21:46:47Z

LWG agrees with this direction; I'll make a few tweaks and then ask for a review.

tkoeppe · 2017-02-05T00:07:18Z

I decided to add a new section "C++ standard library headers [depr.cpp.headers]", as a sibling section to the existing "C standard library headers [depr.c.headers]", and the new section comes first so that the existing one can refer to it.

@zygoloid: Ready for review and application.

@mclow: Could you take a final look to see if this is in accord with LWG guidance?

mclow · 2017-02-05T21:54:25Z

This looks good to me.

AlisdairM reviewed Nov 16, 2016

View reviewed changes

tkoeppe added the lwg Issue must be reviewed by LWG. label Nov 16, 2016

tkoeppe force-pushed the cnew branch 2 times, most recently from 2144255 to 245b1c2 Compare November 17, 2016 14:17

tkoeppe force-pushed the master branch 2 times, most recently from d2a10b4 to c1f9b30 Compare November 17, 2016 14:36

tkoeppe force-pushed the cnew branch 5 times, most recently from 1aa12e4 to 2e444e9 Compare November 18, 2016 00:28

AlisdairM reviewed Nov 18, 2016

View reviewed changes

tkoeppe force-pushed the cnew branch 2 times, most recently from 4cacf3e to 7a820ce Compare November 20, 2016 21:26

tkoeppe force-pushed the cnew branch 4 times, most recently from 806773a to 0105575 Compare November 28, 2016 21:23

jensmaurer mentioned this pull request Dec 1, 2016

remove counts of headers etc from [headers] and [depr.c.headers] #766

Closed

jensmaurer reviewed Dec 1, 2016

View reviewed changes

tkoeppe commented Dec 7, 2016

View reviewed changes

tkoeppe force-pushed the cnew branch 2 times, most recently from 59a6962 to b32a9c1 Compare December 15, 2016 00:09

tkoeppe force-pushed the cnew branch from b32a9c1 to 5b774d2 Compare December 18, 2016 23:36

tkoeppe force-pushed the cnew branch from 5b774d2 to 0643b28 Compare January 13, 2017 12:40

tkoeppe force-pushed the cnew branch from 0643b28 to 3e4f74f Compare February 5, 2017 00:01

jensmaurer removed the lwg Issue must be reviewed by LWG. label Feb 7, 2017

jensmaurer assigned zygoloid Feb 7, 2017

tkoeppe force-pushed the cnew branch 2 times, most recently from b4f390e to 9f28d7c Compare February 23, 2017 00:17

zygoloid added this to the C++17 milestone Mar 3, 2017

tkoeppe force-pushed the cnew branch 2 times, most recently from f464c00 to 826e03a Compare March 16, 2017 01:53

Move specification of deprecated headers to Annex D

331856d

tkoeppe force-pushed the cnew branch from 826e03a to 331856d Compare March 16, 2017 02:05

zygoloid merged commit 8139f2f into cplusplus:master Mar 16, 2017

tkoeppe deleted the cnew branch March 16, 2017 02:15

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Move specification of deprecated headers to Annex D #1054

Move specification of deprecated headers to Annex D #1054

tkoeppe commented Nov 16, 2016

tkoeppe commented Nov 16, 2016

AlisdairM Nov 16, 2016

tkoeppe Nov 16, 2016

AlisdairM Nov 16, 2016

AlisdairM Nov 16, 2016

tkoeppe Nov 16, 2016

AlisdairM Nov 16, 2016

tkoeppe Nov 17, 2016

AlisdairM Nov 16, 2016

tkoeppe Nov 17, 2016

AlisdairM Nov 16, 2016

tkoeppe Nov 16, 2016

AlisdairM Nov 16, 2016

tkoeppe Nov 17, 2016

AlisdairM Nov 18, 2016 •

edited

tkoeppe Nov 18, 2016

jensmaurer Dec 1, 2016

tkoeppe Dec 1, 2016 •

edited

tkoeppe Dec 1, 2016

jensmaurer Dec 7, 2016

tkoeppe Dec 7, 2016

jensmaurer commented Dec 1, 2016

tkoeppe left a comment

tkoeppe Dec 1, 2016

jensmaurer commented Dec 12, 2016

jensmaurer commented Dec 15, 2016

tkoeppe commented Dec 15, 2016

tkoeppe commented Feb 4, 2017

tkoeppe commented Feb 5, 2017

mclow commented Feb 5, 2017

Move specification of deprecated headers to Annex D #1054

Move specification of deprecated headers to Annex D #1054

Conversation

tkoeppe commented Nov 16, 2016

tkoeppe commented Nov 16, 2016

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

AlisdairM Nov 18, 2016 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

tkoeppe Dec 1, 2016 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jensmaurer commented Dec 1, 2016

tkoeppe left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jensmaurer commented Dec 12, 2016

jensmaurer commented Dec 15, 2016

tkoeppe commented Dec 15, 2016

tkoeppe commented Feb 4, 2017

tkoeppe commented Feb 5, 2017

mclow commented Feb 5, 2017

AlisdairM Nov 18, 2016 •

edited

tkoeppe Dec 1, 2016 •

edited