bug#28571: 25.3; Compile error on macOS Sierra

bug#28571: 25.3; Compile error on macOS Sierra

[Philipp]

Compilation fails on macOS Sierra at repository version
c0af83b6ccf2dab9a515dd7f52eb9d4500275ae3:

$ make
/Applications/Xcode.app/Contents/Developer/usr/bin/make -C lib all
  CC       utimens.o
utimens.c:250:20: error: 'utimensat' is only available on macOS 10.13 or newer [-Werror,-Wunguarded-availability-new]
          result = utimensat (AT_FDCWD, file, ts, 0);
                   ^~~~~~~~~
/Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX10.13.sdk/usr/include/sys/stat.h:374:5: note: 'utimensat'
      has been explicitly marked partial here
int     utimensat(int __fd, const char *__path, const struct timespec __times[2],
        ^
utimens.c:250:20: note: enclose 'utimensat' in a __builtin_available check to silence this warning
          result = utimensat (AT_FDCWD, file, ts, 0);
                   ^~~~~~~~~
utimens.c:272:20: error: 'futimens' is only available on macOS 10.13 or newer [-Werror,-Wunguarded-availability-new]
          result = futimens (fd, ts);
                   ^~~~~~~~
/Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX10.13.sdk/usr/include/sys/stat.h:373:5: note: 'futimens'
      has been explicitly marked partial here
int     futimens(int __fd, const struct timespec __times[2]) __API_AVAILABLE(macosx(10.13), ios(11.0), tvos(11.0), watchos(4.0));
        ^
utimens.c:272:20: note: enclose 'futimens' in a __builtin_available check to silence this warning
          result = futimens (fd, ts);
                   ^~~~~~~~
utimens.c:566:16: error: 'utimensat' is only available on macOS 10.13 or newer [-Werror,-Wunguarded-availability-new]
      result = utimensat (AT_FDCWD, file, ts, AT_SYMLINK_NOFOLLOW);
               ^~~~~~~~~
/Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX10.13.sdk/usr/include/sys/stat.h:374:5: note: 'utimensat'
      has been explicitly marked partial here
int     utimensat(int __fd, const char *__path, const struct timespec __times[2],
        ^
utimens.c:566:16: note: enclose 'utimensat' in a __builtin_available check to silence this warning
      result = utimensat (AT_FDCWD, file, ts, AT_SYMLINK_NOFOLLOW);
               ^~~~~~~~~
3 errors generated.
make[1]: *** [utimens.o] Error 1
make: *** [lib] Error 2

$ ./config.status --config
'--enable-checking' '--enable-check-lisp-object-type' '--enable-gtk-deprecation-warnings' '--with-modules' '--without-xml2' '--without-pop' '--with-mailutils' '--enable-gcc-warnings' 'MAKEINFO=/usr/local/opt/texinfo/bin/makeinfo' 'CFLAGS=-O0 -ggdb3' 'LDFLAGS=-O0'



In GNU Emacs 25.3.1 (x86_64-apple-darwin16.7.0, NS appkit-1504.83 Version 10.12.6 (Build 16G29))
 of 2017-09-15 built on p
Windowing system distributor 'Apple', version 10.3.1504
Configured using:
 'configure --disable-dependency-tracking --disable-silent-rules
 --enable-locallisppath=/usr/local/share/emacs/site-lisp
 --infodir=/usr/local/Cellar/emacs/25.3/share/info/emacs
 --prefix=/usr/local/Cellar/emacs/25.3 --without-x --with-xml2
 --without-dbus --without-gnutls --without-imagemagick --with-modules
 --with-ns --disable-ns-self-contained'

Configured features:
JPEG NOTIFY ACL LIBXML2 ZLIB TOOLKIT_SCROLL_BARS NS MODULES

Important settings:
  value of $LANG: de_DE.UTF-8
  locale-coding-system: utf-8-unix

Major mode: Lisp Interaction

Minor modes in effect:
  tooltip-mode: t
  global-eldoc-mode: t
  electric-indent-mode: t
  mouse-wheel-mode: t
  tool-bar-mode: t
  menu-bar-mode: t
  file-name-shadow-mode: t
  global-font-lock-mode: t
  font-lock-mode: t
  blink-cursor-mode: t
  auto-composition-mode: t
  auto-encryption-mode: t
  auto-compression-mode: t
  line-number-mode: t
  transient-mark-mode: t

Recent messages:
For information about GNU Emacs and the GNU system, type C-h C-a.

Load-path shadows:
None found.

Features:
(shadow sort mail-extr emacsbug message dired format-spec rfc822 mml
mml-sec password-cache epg epg-config gnus-util mm-decode mm-bodies
mm-encode mail-parse rfc2231 mailabbrev gmm-utils mailheader sendmail
rfc2047 rfc2045 ietf-drums mm-util help-fns help-mode easymenu
cl-loaddefs pcase cl-lib mail-prsvr mail-utils time-date mule-util
tooltip eldoc electric uniquify ediff-hook vc-hooks lisp-float-type
mwheel ns-win ucs-normalize term/common-win tool-bar dnd fontset image
regexp-opt fringe tabulated-list newcomment elisp-mode lisp-mode
prog-mode register page menu-bar rfn-eshadow timer select scroll-bar
mouse jit-lock font-lock syntax facemenu font-core frame cl-generic cham
georgian utf-8-lang misc-lang vietnamese tibetan thai tai-viet lao
korean japanese eucjp-ms cp51932 hebrew greek romanian slovak czech
european ethiopic indian cyrillic chinese charscript case-table epa-hook
jka-cmpr-hook help simple abbrev minibuffer cl-preloaded nadvice
loaddefs button faces cus-face macroexp files text-properties overlay
sha1 md5 base64 format env code-pages mule custom widget
hashtable-print-readable backquote kqueue cocoa ns multi-tty
make-network-process emacs)

Memory information:
((conses 16 196024 7412)
 (symbols 48 19650 0)
 (miscs 40 50 187)
 (strings 32 15009 5265)
 (string-bytes 1 435439)
 (vectors 16 32913)
 (vector-slots 8 653955 6213)
 (floats 8 160 12)
 (intervals 56 210 0)
 (buffers 976 18))

bug#28571: 25.3; Compile error on macOS Sierra

[Alan Third]

On Sat, Sep 23, 2017 at 09:08:21PM +0200, Philipp wrote:
> 
> Compilation fails on macOS Sierra at repository version
> c0af83b6ccf2dab9a515dd7f52eb9d4500275ae3:
> 
> $ make
> /Applications/Xcode.app/Contents/Developer/usr/bin/make -C lib all
>   CC       utimens.o
> utimens.c:250:20: error: 'utimensat' is only available on macOS 10.13 or newer [-Werror,-Wunguarded-availability-new]
>           result = utimensat (AT_FDCWD, file, ts, 0);
>                    ^~~~~~~~~

I can’t replicate, but I don’t have the full xcode installed, just the
command line stuff.

I take it you upgraded Xcode the other day? (To version 9, I think?)

Is there maybe a test in configure that’s incorrectly identifying this
utimens.c stuff as OK?
-- 
Alan Third

bug#28571: 25.3; Compile error on macOS Sierra

[Philipp Stephani]

[-- Attachment #1: Type: text/plain, Size: 1151 bytes --]

Alan Third <alan@idiocy.org> schrieb am Sa., 23. Sep. 2017 um 21:31 Uhr:

> On Sat, Sep 23, 2017 at 09:08:21PM +0200, Philipp wrote:
> >
> > Compilation fails on macOS Sierra at repository version
> > c0af83b6ccf2dab9a515dd7f52eb9d4500275ae3:
> >
> > $ make
> > /Applications/Xcode.app/Contents/Developer/usr/bin/make -C lib all
> >   CC       utimens.o
> > utimens.c:250:20: error: 'utimensat' is only available on macOS 10.13 or
> newer [-Werror,-Wunguarded-availability-new]
> >           result = utimensat (AT_FDCWD, file, ts, 0);
> >                    ^~~~~~~~~
>
> I can’t replicate, but I don’t have the full xcode installed, just the
> command line stuff.
>
> I take it you upgraded Xcode the other day? (To version 9, I think?)
>

Yes, indeed.


>
> Is there maybe a test in configure that’s incorrectly identifying this
> utimens.c stuff as OK?
>
>
I do have configure output like this:

checking for futimens... yes
checking for utimensat... yes

So I think the functions are there, but the compilation shouldn't use them
since they would make the binary incompatible with old versions of macOS.

[-- Attachment #2: Type: text/html, Size: 1783 bytes --]

bug#28571: 25.3; Compile error on macOS Sierra

[Alan Third]

[-- Attachment #1: Type: text/plain, Size: 721 bytes --]

On Sun, Sep 24, 2017 at 07:49:55AM +0000, Philipp Stephani wrote:
> Alan Third <alan@idiocy.org> schrieb am Sa., 23. Sep. 2017 um 21:31 Uhr:
> >
> > Is there maybe a test in configure that’s incorrectly identifying this
> > utimens.c stuff as OK?
> >
> >
> I do have configure output like this:
> 
> checking for futimens... yes
> checking for utimensat... yes
> 
> So I think the functions are there, but the compilation shouldn't use them
> since they would make the binary incompatible with old versions of macOS.

I agree. I’ve attached a patch that should sort it.

I’m not 100% sure that DARWIN_OS is the right define to check for, but
it seems more appropriate than any others I can think of.
-- 
Alan Third

[-- Attachment #2: 0001-Fix-build-on-macOS-bug-28571.patch --]
[-- Type: text/plain, Size: 844 bytes --]

From d77a2e45004f5d6d8496e5fed260774205efe42a Mon Sep 17 00:00:00 2001
From: Alan Third <alan@idiocy.org>
Date: Sun, 24 Sep 2017 12:01:03 +0100
Subject: [PATCH] Fix build on macOS (bug#28571)

* src/conf_post.h (HAVE_FUTIMENS, HAVE_FUTIMESAT) [DARWIN_OS]:
Undefine.
---
 src/conf_post.h | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/src/conf_post.h b/src/conf_post.h
index febdb8b8bf..dfdb0aa3eb 100644
--- a/src/conf_post.h
+++ b/src/conf_post.h
@@ -397,3 +397,11 @@ extern int emacs_setenv_TZ (char const *);
 #else
 # define UNINIT /* empty */
 #endif
+
+/* macOS 10.13 supports futimens and futimesat, older versions don't
+   but can appear as though they do.  Disable them entirely to avoid
+   breaking cross-version builds on macOS.  */
+#ifdef DARWIN_OS
+# undef HAVE_FUTIMENS
+# undef HAVE_FUTIMESAT
+#endif
-- 
2.14.1

bug#28571: 25.3; Compile error on macOS Sierra

[Philipp Stephani]

[-- Attachment #1: Type: text/plain, Size: 1174 bytes --]

Alan Third <alan@idiocy.org> schrieb am So., 24. Sep. 2017 um 13:07 Uhr:

> On Sun, Sep 24, 2017 at 07:49:55AM +0000, Philipp Stephani wrote:
> > Alan Third <alan@idiocy.org> schrieb am Sa., 23. Sep. 2017 um 21:31 Uhr:
> > >
> > > Is there maybe a test in configure that’s incorrectly identifying this
> > > utimens.c stuff as OK?
> > >
> > >
> > I do have configure output like this:
> >
> > checking for futimens... yes
> > checking for utimensat... yes
> >
> > So I think the functions are there, but the compilation shouldn't use
> them
> > since they would make the binary incompatible with old versions of macOS.
>
> I agree. I’ve attached a patch that should sort it.
>

Thanks, the patch works, but you also need to #undef HAVE_UTIMENSAT.
However, I'm wondering whether that's the right way to solve it. It means
we're making some of the checks manual that have previously been automated,
which implies that we need to recheck in the future whether the checks are
still appropriate etc. Wouldn't it be better to modify configure.ac so that
it doesn't detect these functions as present in the first place if we don't
want to use them?

[-- Attachment #2: Type: text/html, Size: 1643 bytes --]

bug#28571: 25.3; Compile error on macOS Sierra

[Alan Third]

On Sun, Sep 24, 2017 at 11:23:47AM +0000, Philipp Stephani wrote:
> Alan Third <alan@idiocy.org> schrieb am So., 24. Sep. 2017 um 13:07 Uhr:
> 
> > On Sun, Sep 24, 2017 at 07:49:55AM +0000, Philipp Stephani wrote:
> > > So I think the functions are there, but the compilation
> > > shouldn't use them since they would make the binary incompatible
> > > with old versions of macOS.
> >
> > I agree. I’ve attached a patch that should sort it.
> >
> 
> Thanks, the patch works, but you also need to #undef HAVE_UTIMENSAT.
> However, I'm wondering whether that's the right way to solve it. It
> means we're making some of the checks manual that have previously
> been automated, which implies that we need to recheck in the future
> whether the checks are still appropriate etc. Wouldn't it be better
> to modify configure.ac so that it doesn't detect these functions as
> present in the first place if we don't want to use them?

I can’t alter the checks directly as they’re part of gnulib (I think).
We’d have to suppress them. But that would be done by putting in a
simple check, ‘is this macOS? If yes, don’t use these functions’,
which is, IMO, no different than what the patch does.

We could get the gnulib people to fix them, but even then we’d still
want to just suppress them as we don’t want to use these functions on
macOS even if they are available due to cross‐version concerns.

If you just think the logic should be in configure.ac rather than
conf_post.h, then I think we might need to get help with how to
suppress or ignore the checks. Unless you know how.
-- 
Alan Third

bug#28571: 25.3; Compile error on macOS Sierra

[YAMAMOTO Mitsuharu]

>>>>> On Sun, 24 Sep 2017 07:49:55 +0000, Philipp Stephani <p.stephani2@gmail.com> said:

>> I take it you upgraded Xcode the other day? (To version 9, I
>> think?)
>> 

> Yes, indeed.

Probably You need to execute "$ xcode-select --install" to install
Command Line Tools for Xcode.  You can actually go without it by
tweaking some environment variables and/or compiler arguments, but I
wouldn't recommend that unless you have a shortage of disk space and
you really understand what you are doing.

				     YAMAMOTO Mitsuharu
				mituharu@math.s.chiba-u.ac.jp

bug#28571: 25.3; Compile error on macOS Sierra

[Alan Third]

Patch pushed to emacs-26.
-- 
Alan Third

bug#28571: 25.3; Compile error on macOS Sierra

[mituharu]

> Patch pushed to emacs-26.

???  The OP did not reply whether he installed the Command Line Tools.
I still think the failure is due to not installing them.

As for binary compatibility issue, futimens, futimesat, utimensat are not
the only such functions.
For example, mkostemp, fstatat, fdopendir, readlinkat, and faccessat are
available on macOS 10.12, but not on all older OS X versions that Emacs 26
supports.
If you want to build a binary on a new version so that it also runs on
older versions,
then you need to use config.status generated on the oldest version anyway.

				     YAMAMOTO Mitsuharu
				mituharu@math.s.chiba-u.ac.jp

bug#28571: 25.3; Compile error on macOS Sierra

[Alan Third]

On Fri, Sep 29, 2017 at 05:58:27AM +0900, mituharu@math.s.chiba-u.ac.jp wrote:
> > Patch pushed to emacs-26.
> 
> ???  The OP did not reply whether he installed the Command Line Tools.
> I still think the failure is due to not installing them.

Sorry, I missed that. I’ve reverted the change.

> As for binary compatibility issue, futimens, futimesat, utimensat are not
> the only such functions.
> For example, mkostemp, fstatat, fdopendir, readlinkat, and faccessat are
> available on macOS 10.12, but not on all older OS X versions that Emacs 26
> supports.
> If you want to build a binary on a new version so that it also runs on
> older versions,
> then you need to use config.status generated on the oldest version anyway.

Thanks for this information. There was a discussion a little while ago
about how to implement a single binary that works across macOS
versions and nobody thought of this particular issue.
-- 
Alan Third

bug#28571: 25.3; Compile error on macOS Sierra

[Eli Zaretskii]

> Date: Sun, 24 Sep 2017 12:06:58 +0100
> From: Alan Third <alan@idiocy.org>
> Cc: 28571@debbugs.gnu.org
> 
> diff --git a/src/conf_post.h b/src/conf_post.h
> index febdb8b8bf..dfdb0aa3eb 100644
> --- a/src/conf_post.h
> +++ b/src/conf_post.h
> @@ -397,3 +397,11 @@ extern int emacs_setenv_TZ (char const *);
>  #else
>  # define UNINIT /* empty */
>  #endif
> +
> +/* macOS 10.13 supports futimens and futimesat, older versions don't
> +   but can appear as though they do.  Disable them entirely to avoid
> +   breaking cross-version builds on macOS.  */
> +#ifdef DARWIN_OS
> +# undef HAVE_FUTIMENS
> +# undef HAVE_FUTIMESAT
> +#endif

I think it's preferable to have this stuff localized to the configure
script.  Having the same issue decided in tow different places makes
this harder to maintain.  Can you add an equivalent test to
configure.ac?

Thanks.

bug#28571: 25.3; Compile error on macOS Sierra

[Philipp Stephani]

[-- Attachment #1: Type: text/plain, Size: 810 bytes --]

YAMAMOTO Mitsuharu <mituharu@math.s.chiba-u.ac.jp> schrieb am Mo., 25. Sep.
2017 um 02:29 Uhr:

> >>>>> On Sun, 24 Sep 2017 07:49:55 +0000, Philipp Stephani <
> p.stephani2@gmail.com> said:
>
> >> I take it you upgraded Xcode the other day? (To version 9, I
> >> think?)
> >>
>
> > Yes, indeed.
>
> Probably You need to execute "$ xcode-select --install" to install
> Command Line Tools for Xcode.  You can actually go without it by
> tweaking some environment variables and/or compiler arguments, but I
> wouldn't recommend that unless you have a shortage of disk space and
> you really understand what you are doing.
>
>
OK, the warning is indeed gone after that, thanks.
Should we put something in configure.ac that checks for this? It's a bit
subtle because compilation almost completely works without it.

[-- Attachment #2: Type: text/html, Size: 1298 bytes --]

bug#28571: 25.3; Compile error on macOS Sierra

[YAMAMOTO Mitsuharu]

>>>>> On Sat, 30 Sep 2017 19:17:33 +0000, Philipp Stephani <p.stephani2@gmail.com> said:

>> Probably You need to execute "$ xcode-select --install" to install
>> Command Line Tools for Xcode.  You can actually go without it by
>> tweaking some environment variables and/or compiler arguments, but
>> I wouldn't recommend that unless you have a shortage of disk space
>> and you really understand what you are doing.
>> 
>> 
> OK, the warning is indeed gone after that, thanks.  Should we put
> something in configure.ac that checks for this? It's a bit subtle
> because compilation almost completely works without it.

Installation of the Command Line Tools is not a strict requirement.
For example, the Nix package manager does not require Xcode but use
its own SDK, IIUC.  On the other hand, MacPorts and Homebrew
explicitly require users to install the Command Line Tools:

  https://www.macports.org/install.php
  https://docs.brew.sh/Installation.html

I guess nowadays most users use some package manager.  If they are
following its installation instructions correctly, they are in a
correct configuration.  Otherwise, they will face another problem
sooner or later anyway.

			     YAMAMOTO Mitsuharu
			mituharu@math.s.chiba-u.ac.jp