Last commit breaks macOS build

Last commit breaks macOS build

[Angelo Graziosi]

This commit: 44c588a25ce231ce05fb535cd6d7162e91214f45 (Port recent 
autogen.sh changes to Darwin) breaks the build on macOS:

./autogen.sh
[...]
Your system has the required tools.
Interfering nt/gnulib.mk from lib/gnulib.mk ...
sed: 46: "This is of m...": command c expects \ followed by text

and the 'configure' is not created...

Angelo

Re: Last commit breaks macOS build

[Paul Eggert]

[-- Attachment #1: Type: text/plain, Size: 538 bytes --]

Angelo Graziosi wrote:

> Interfering nt/gnulib.mk from lib/gnulib.mk ...

Did it really say "Interfering"? :-)

> sed: 46: "This is of m...": command c expects \ followed by text
>
> and the 'configure' is not created...

This is caused by our recent attempt to better-automate MS-Windows builds. I 
fixed one nonportable construct, but evidently there are others.

It's time to break out a bigger hammer. Please try the attached patch. I have 
tested it on Fedora, but not on MS-Windows and so have not installed it yet.

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: 0001-Use-GNU-Make-instead-of-sed-Automake-for-nt-lib.patch --]
[-- Type: text/x-diff; name="0001-Use-GNU-Make-instead-of-sed-Automake-for-nt-lib.patch", Size: 10602 bytes --]

From d59053569fb37f80487810f6e18da0ac9030a0da Mon Sep 17 00:00:00 2001
From: Paul Eggert <eggert@cs.ucla.edu>
Date: Wed, 4 Jan 2017 19:13:14 -0800
Subject: [PATCH] Use GNU Make instead of sed+Automake for nt lib

Problem reported by Angelo Graziosi.  Instead of continuing to
hack something in autogen.sh, which is a pain to port and which
is a prepass to Automake, this patch uses GNU Make instead,
so that there is no need to have anything in autogen.sh and
no need for an nt/gnulib.mk file.
* .gitignore: Remove nt/gnulib.mk.
* Makefile.in (gnulib_mk, $(srcdir)/nt/gnulib.mk): Remove.
(AUTOMAKE_INPUTS): Use $(srcdir)/lib/gnulib.mk, not gnulib_mk.
* autogen.sh: Do not build nt/gnulib.mk.
* configure.ac (GNULIB_MK): Remove.
(SYSTEM_TYPE): Subst.
(MT): New macro.  Subst it.
* lib/Makefile.am: Do not include ../nt/gnulib.mk.
Use $(SYSTEM_TYPE), not BUILDING_FOR_WINDOWSNT.
(libgnu_a_SOURCES_not, EXTRA_libgnu_a_SOURCES_not): New macros.
(libgnu_a_SOURCES, EXTRA_libgnu_a_SOURCES) [windows-nt]:
Filter out unwanted modules.
* make-dist: Omit removed files.
* nt/Makefile.in (AM_V_GEN, am__v_GEN_, am__v_GEN_0)
(am__v_GEN_1, ${srcdir}/gnulib.mk): Remove.
* nt/gnulib-modules-to-delete.cfg: Remove.
---
 .gitignore                      |  1 -
 Makefile.in                     |  8 +----
 autogen.sh                      | 10 +------
 configure.ac                    | 22 ++++----------
 lib/Makefile.am                 | 19 ++++++++----
 make-dist                       |  2 +-
 nt/Makefile.in                  | 10 -------
 nt/gnulib-modules-to-delete.cfg | 66 -----------------------------------------
 8 files changed, 23 insertions(+), 115 deletions(-)
 delete mode 100644 nt/gnulib-modules-to-delete.cfg

diff --git a/.gitignore b/.gitignore
index e8eb4fd..85c06c9 100644
--- a/.gitignore
+++ b/.gitignore
@@ -288,6 +288,5 @@ lib/SYS/
 /site-lisp/
 nt/emacs.rc
 nt/emacsclient.rc
-nt/gnulib.mk
 src/gdb.ini
 /var/
diff --git a/Makefile.in b/Makefile.in
index b9aaf5b..3cb1f5e 100644
--- a/Makefile.in
+++ b/Makefile.in
@@ -98,9 +98,6 @@ configuration=
 ### The nt/ subdirectory gets built only for MinGW
 NTDIR=@NTDIR@
 
-### $(srcdir)/nt/gnulib.mk for MinGW, else $(srcdir)/lib/gnulib.mk
-gnulib_mk=@GNULIB_MK@
-
 # 'make' verbosity.
 AM_DEFAULT_VERBOSITY = @AM_DEFAULT_VERBOSITY@
 
@@ -454,13 +451,10 @@ $(srcdir)/aclocal.m4:
 	cd $(srcdir) && ACLOCAL_PATH='$(ACLOCAL_PATH)' $(ACLOCAL) -I m4
 
 AUTOMAKE_INPUTS = $(srcdir)/aclocal.m4 $(srcdir)/lib/Makefile.am \
-  $(gnulib_mk)
+  $(srcdir)/lib/gnulib.mk
 $(srcdir)/lib/Makefile.in: $(AUTOMAKE_INPUTS)
 	cd $(srcdir) && $(AUTOMAKE) --gnu -a -c lib/Makefile
 
-$(srcdir)/nt/gnulib.mk: $(srcdir)/lib/Makefile.in
-	$(MAKE) -C $(srcdir)/nt gnulib.mk
-
 # Regenerate files that this makefile would have made, if this makefile
 # had been built by Automake.  The name 'am--refresh' is for
 # compatibility with subsidiary Automake-generated makefiles.
diff --git a/autogen.sh b/autogen.sh
index f6bfde6..7381bc3 100755
--- a/autogen.sh
+++ b/autogen.sh
@@ -220,17 +220,9 @@ do_git=
   fi
 
   echo 'Your system has the required tools.'
-
-  ## Create nt/gnulib.mk if it doesn't exist, as autoreconf will need it.
-  if test ! -f nt/gnulib.mk; then
-      echo 'Inferring nt/gnulib.mk from lib/gnulib.mk ...'
-      metascript='/^[^#]/s|^.*$|/^## begin  *gnulib module &/,/^## end  *gnulib module &/c ## gnulib module & removed|'
-      script=`sed "$metascript" nt/gnulib-modules-to-delete.cfg` || exit
-      sed "$script" lib/gnulib.mk > nt/gnulib.mk || exit
-  fi
-
   echo "Running 'autoreconf -fi -I m4' ..."
 
+
   ## Let autoreconf figure out what, if anything, needs doing.
   ## Use autoreconf's -f option in case autoreconf itself has changed.
   autoreconf -fi -I m4 || exit $?
diff --git a/configure.ac b/configure.ac
index 262f367..7fd4988 100644
--- a/configure.ac
+++ b/configure.ac
@@ -771,22 +771,6 @@ AC_DEFUN
   test "x$GCC_TEST_OPTIONS" != x && CC="$CC $GCC_TEST_OPTIONS"
 fi
 
-dnl This is used in lib/Makefile.am to use nt/gnulib.mk, the
-dnl alternative to lib/gnulib.mk, so as to avoid generating header files
-dnl that clash with MinGW.
-case $opsys in
-  mingw32 )
-    GNULIB_MK='$(srcdir)/nt/gnulib.mk'
-    ;;
-  * )
-    GNULIB_MK='$(srcdir)/lib/gnulib.mk'
-    ;;
-
-esac
-AC_SUBST([GNULIB_MK])
-
-AM_CONDITIONAL([BUILDING_FOR_WINDOWSNT], [test "x$opsys" = "xmingw32"])
-
 # Avoid gnulib's tests for -lcrypto, so that there's no static dependency on it.
 AC_DEFUN([gl_CRYPTO_CHECK])
 # Avoid gnulib's tests for HAVE_WORKING_O_NOATIME and HAVE_WORKING_O_NOFOLLOW,
@@ -1555,7 +1539,13 @@ AC_DEFUN
 AC_SUBST(LIB_MATH)
 AC_DEFINE_UNQUOTED(SYSTEM_TYPE, "$SYSTEM_TYPE",
   [The type of system you are compiling for; sets 'system-type'.])
+AC_SUBST([SYSTEM_TYPE])
 
+# The empty (MT) string.
+# Use this in Makefile.am before GNU Make conditionals;
+# otherwise, Automake gets confused.
+MT=
+AC_SUBST([MT])
 
 pre_PKG_CONFIG_CFLAGS=$CFLAGS
 pre_PKG_CONFIG_LIBS=$LIBS
diff --git a/lib/Makefile.am b/lib/Makefile.am
index e6d90dd..141838c 100644
--- a/lib/Makefile.am
+++ b/lib/Makefile.am
@@ -29,13 +29,22 @@ SUFFIXES =
 AM_CFLAGS = $(PROFILING_CFLAGS) $(GNULIB_WARN_CFLAGS) $(WERROR_CFLAGS)
 DEFAULT_INCLUDES = -I. -I$(top_srcdir)/lib -I../src -I$(top_srcdir)/src
 
-if BUILDING_FOR_WINDOWSNT
-include ../nt/gnulib.mk
-else
 include gnulib.mk
 
-libgnu_a_SOURCES += openat-die.c save-cwd.c
-endif
+@MT@ifeq ($(SYSTEM_TYPE), windows-nt)
+@MT@  libgnu_a_SOURCES_not = \
+@MT@    acl-errno-valid.c acl-internal.c allocator.c \
+@MT@    careadlinkat.c get-permissions.c pipe2.c \
+@MT@    set-permissions.c tempname.c unistd.c
+@MT@  libgnu_a_SOURCES := \
+@MT@    $(filter-out $(libgnu_a_SOURCES_not), $(libgnu_a_SOURCES))
+@MT@  EXTRA_libgnu_a_SOURCES_not = \
+@MT@    acl_entries.c dirfd.c fcntl.c mkostemp.c openat-proc.c secure_getenv.c
+@MT@  EXTRA_libgnu_a_SOURCES := \
+@MT@    $(filter-out $(EXTRA_libgnu_a_SOURCES_not), $(EXTRA_libgnu_a_SOURCES))
+@MT@else
+@MT@  libgnu_a_SOURCES += openat-die.c save-cwd.c
+@MT@endif
 
 if HYBRID_MALLOC_LIB
 noinst_LIBRARIES += libegnu.a
diff --git a/make-dist b/make-dist
index 41203b2..2d1a218 100755
--- a/make-dist
+++ b/make-dist
@@ -482,7 +482,7 @@ files=
 (cd nt
  ln emacs-x86.manifest emacs-x64.manifest ../${tempdir}/nt
  ln [a-z]*.bat [a-z]*.[ch] ../${tempdir}/nt
- ln *.in gnulib.mk gnulib-modules-to-delete.cfg ../${tempdir}/nt
+ ln *.in ../${tempdir}/nt
  ln mingw-cfg.site epaths.nt INSTALL.W64 ../${tempdir}/nt
  ln ChangeLog.*[0-9] INSTALL README README.W32 ../${tempdir}/nt)
 
diff --git a/nt/Makefile.in b/nt/Makefile.in
index 89f4460..7e911db 100644
--- a/nt/Makefile.in
+++ b/nt/Makefile.in
@@ -54,11 +54,6 @@ am__v_CCLD_ =
 am__v_CCLD_0 = @echo "  CCLD    " $@;
 am__v_CCLD_1 =
 
-AM_V_GEN = $(am__v_GEN_@AM_V@)
-am__v_GEN_ = $(am__v_GEN_@AM_DEFAULT_V@)
-am__v_GEN_0 = @echo "  GEN     " $@;
-am__v_GEN_1 =
-
 AM_V_RC = $(am__v_RC_@AM_V@)
 am__v_RC_ = $(am__v_RC_@AM_DEFAULT_V@)
 am__v_RC_0 = @echo "  RC      " $@;
@@ -260,8 +255,3 @@ runemacs${EXEEXT}:
 emacs.res ../src/emacs.res: emacs.rc ${srcdir}/icons/emacs.ico \
   ${srcdir}/icons/hand.cur ${srcdir}/$(EMACS_MANIFEST)
 	$(AM_V_RC)${WINDRES} -I ${srcdir} -O coff -o $@ $<
-
-${srcdir}/gnulib.mk: ${srcdir}/gnulib-modules-to-delete.cfg ${srcdir}/../lib/gnulib.mk
-	$(AM_V_GEN)sed '/^[^#]/s|^.*$$|/^## begin  *gnulib module &/,/^## end  *gnulib module &/c ## gnulib module & removed|' < $< | \
-	  sed -f- ${srcdir}/../lib/gnulib.mk > $@-t && \
-	  ${srcdir}/../build-aux/move-if-change $@-t $@
diff --git a/nt/gnulib-modules-to-delete.cfg b/nt/gnulib-modules-to-delete.cfg
deleted file mode 100644
index 09bea0e..0000000
--- a/nt/gnulib-modules-to-delete.cfg
+++ /dev/null
@@ -1,66 +0,0 @@
-# This is the list of modules to be deleted from ../lib/gnulib.mk.
-#
-# We delete them because they either conflict with MinGW headers or
-# headers in nt/inc, or because those modules won't compile with
-# MinGW, or because Emacs already has the corresponding facilities as
-# part of Emacs sources, and their implementation is incompatible with
-# Gnulib's.
-#
-# In general, do NOT remove anything from ../lib/gnulib.mk that
-# doesn't need to be removed, to minimize the differences from
-# upstream gnulib.mk and thus make the maintenance easier.  Every
-# header file whose generation is controlled by configure-time tests
-# does NOT need to be removed; instead, force the configure script to
-# accept whatever MinGW has to offer, by defining the appropriate
-# Autoconf variable in the nt/mingw-cfg.site file.  Headers that are
-# generated conditionally have the tell-tale "if GL_GENERATE_foo_H"
-# condition before their Makefile snippet in this file.  Likewise, do
-# NOT remove gnulib modules which introduce header files that don't
-# exist in MinGW and in nt/inc/, since they cannot possibly clash
-# with anything.  Gnulib modules that introduce source *.c files also
-# need not be removed; if they define functions that could clash with
-# the w32 substitutes in Emacs, disable their compilation by defining
-# suitable variables in nt/mingw-cfg.site.
-# ----------------------------------------------------------------------
-#
-# Copyright (C) 2017 Free Software Foundation, Inc.
-#
-# This file is free software; you can redistribute it and/or modify
-# it under the terms of the GNU General Public License as published by
-# the Free Software Foundation; either version 3 of the License, or
-# (at your option) any later version.
-#
-# This file is distributed in the hope that it will be useful,
-# but WITHOUT ANY WARRANTY; without even the implied warranty of
-# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-# GNU General Public License for more details.
-#
-# You should have received a copy of the GNU General Public License
-# along with this file.  If not, see <http://www.gnu.org/licenses/>.
-#
-# As a special exception to the GNU General Public License,
-# this file may be distributed as part of a program that
-# contains a configuration script generated by Autoconf, under
-# the same distribution terms as the rest of that program.
-#
-acl-permissions
-allocator
-at-internal
-careadlinkat
-dirent
-dirfd
-fcntl
-fcntl-h
-inttypes-incomplete
-mkostemp
-pipe2
-secure_getenv
-signal-h
-stdio
-stdlib
-sys_select
-sys_stat
-sys_time
-sys_types
-tempname
-unistd
-- 
2.7.4

Re: Last commit breaks macOS build

[Richard Copley]

On 5 January 2017 at 03:21, Paul Eggert <eggert@cs.ucla.edu> wrote:
> Angelo Graziosi wrote:
>
>> Interfering nt/gnulib.mk from lib/gnulib.mk ...
>
>
> Did it really say "Interfering"? :-)
>
>> sed: 46: "This is of m...": command c expects \ followed by text
>>
>> and the 'configure' is not created...
>
>
> This is caused by our recent attempt to better-automate MS-Windows builds. I
> fixed one nonportable construct, but evidently there are others.
>
> It's time to break out a bigger hammer. Please try the attached patch. I
> have tested it on Fedora, but not on MS-Windows and so have not installed it
> yet.

Hi Paul,
In MSYS2 on Windows, with the patch I get this error:

gcc  -DHAVE_CONFIG_H -I. -I../../repo/emacs/lib -I../src
-I../../repo/emacs/src   -march=native -mtune=native
-D_WIN32_WINNT=_WIN32_WINNT_WIN7 -DUSE_CRT_DLL=1 -I
/g/emacs/build/../repo/emacs/nt/inc  -fno-common /* -W... */ -Og -g
-ggdb -MT set-permissions.o -MD -MP -MF .deps/set-permissions.Tpo -c
-o set-permissions.o ../../repo/emacs/lib/set-permissions.c
../../repo/emacs/lib/set-permissions.c: In function 'set_acls':
../../repo/emacs/lib/set-permissions.c:500:6: error: #error Must have
acl_delete_def_file (see POSIX 1003.1e draft 17).
 #    error Must have acl_delete_def_file (see POSIX 1003.1e draft 17).
      ^~~~~
../../repo/emacs/lib/set-permissions.c:539:13: warning: implicit
declaration of function 'acl_delete_def_file'
[-Wimplicit-function-declaration]
       ret = acl_delete_def_file (name);
             ^~~~~~~~~~~~~~~~~~~
../../repo/emacs/lib/set-permissions.c:539:7: warning: nested extern
declaration of 'acl_delete_def_file' [-Wnested-externs]
       ret = acl_delete_def_file (name);
       ^~~

Without the patch, set-permissions.c isn't compiled.

Re: Last commit breaks macOS build

[Angelo Graziosi]

Hi Paul,

really I spoke of macOS build not MSWINDOWS (MSW).. But now, as Richard 
reported,  it seems that also MSW build is broken...


In any case


> On 5 January 2017 at 03:21, Paul Eggert
>> Did it really say "Interfering"? :-)

Yes, I copy/pasted it..

You should consider that macOS does not use GNU sed but another version 
(unix/BSD?), and what works on GNU/Linux not always works in the same 
manne on macOS...

Angelo

Re: Last commit breaks macOS build

[Eli Zaretskii]

> From: Angelo Graziosi <angelo.graziosi@alice.it>
> Date: Thu, 5 Jan 2017 10:54:16 +0100
> Cc: Emacs developers <emacs-devel@gnu.org>
> 
> Hi Paul,
> 
> really I spoke of macOS build not MSWINDOWS (MSW).. But now, as Richard 
> reported,  it seems that also MSW build is broken...

No, it's not broken because Paul didn't install his changes.  Richard
was just giving him a heads-up.  (Thanks, Richard!)

Re: Last commit breaks macOS build

[Eli Zaretskii]

> From: Angelo Graziosi <angelo.graziosi@alice.it>
> Date: Thu, 5 Jan 2017 10:54:16 +0100
> Cc: Emacs developers <emacs-devel@gnu.org>
> 
> You should consider that macOS does not use GNU sed but another version 
> (unix/BSD?), and what works on GNU/Linux not always works in the same 
> manne on macOS...

That's my fault, not Paul's.  Sorry about that.

Could you please try the current master?  I attempted to fix this
problem.

Re: Last commit breaks macOS build

[Eli Zaretskii]

> From: Paul Eggert <eggert@cs.ucla.edu>
> Date: Wed, 4 Jan 2017 19:21:11 -0800
> 
> This is caused by our recent attempt to better-automate MS-Windows builds. I 
> fixed one nonportable construct, but evidently there are others.

Yes, sorry about that.  Mea culpa.

> It's time to break out a bigger hammer. Please try the attached patch. I have 
> tested it on Fedora, but not on MS-Windows and so have not installed it yet.

I'd prefer to avoid the need for the people who maintain the
MS-Windows build to understand the structure of lib/gnulib.mk in too
many details, such as what files comprise each Gnulib package, and
what is the meaning of the likes of EXTRA_libgnu_a_SOURCES.  OTOH, I'd
prefer not to bother people who merge from Gnulib with understanding
what parts the MS-Windows build needs to avoid.

Having a simple list of Gnulib packages to remove from lib/gnulib.mk,
similar to nt/gnulib-modules-to-delete.cfg, is therefore preferable to
a much more detailed list that you propose to maintain manually in
lib/Makefile.am.  It is also simpler, since the syntax of Makefile.am
doesn't need to be familiar to whoever modifies the list of modules.

Is it possible to use a similar plain list of modules while solving
this with GNU Make features?  If not, I think we should stick with the
current solution.  It's not a bad one, I think, once we clean up the
portability issues.  (I just tried to solve a couple of such issues.)

Thanks.

Re: Last commit breaks macOS build

[Angelo Graziosi]

Il 05/01/2017 17:16, Eli Zaretskii ha scritto:
>> From: Angelo Graziosi <angelo.graziosi@alice.it>
>> Date: Thu, 5 Jan 2017 10:54:16 +0100
>> Cc: Emacs developers <emacs-devel@gnu.org>
>>
>> You should consider that macOS does not use GNU sed but another version
>> (unix/BSD?), and what works on GNU/Linux not always works in the same
>> manne on macOS...
>
> That's my fault, not Paul's.  Sorry about that.
>
> Could you please try the current master?  I attempted to fix this
> problem.
>

On macOS same issue with master-003954e67fbe97c6ac786983912f572c48a331f5.

Maybe this can help: 
http://lists.gnu.org/archive/html/emacs-devel/2017-01/msg00071.html

Re: Last commit breaks macOS build

[Eli Zaretskii]

> Cc: rcopley@gmail.com, eggert@cs.ucla.edu, emacs-devel@gnu.org
> From: Angelo Graziosi <angelo.graziosi@alice.it>
> Date: Thu, 5 Jan 2017 17:37:00 +0100
> 
> > Could you please try the current master?  I attempted to fix this
> > problem.
> >
> 
> On macOS same issue with master-003954e67fbe97c6ac786983912f572c48a331f5.

Could you please show the error messages you see with the latest
master?


> Maybe this can help: 
> http://lists.gnu.org/archive/html/emacs-devel/2017-01/msg00071.html

That didn't work on other systems, so we cannot use it as-is.

Thanks.

Re: Last commit breaks macOS build

[Angelo Graziosi]

Il 05/01/2017 18:04, Eli Zaretskii ha scritto:
>> Cc: rcopley@gmail.com, eggert@cs.ucla.edu, emacs-devel@gnu.org
>> From: Angelo Graziosi <angelo.graziosi@alice.it>
>> Date: Thu, 5 Jan 2017 17:37:00 +0100
>>
>>> Could you please try the current master?  I attempted to fix this
>>> problem.
>>>
>>
>> On macOS same issue with master-003954e67fbe97c6ac786983912f572c48a331f5.
>
> Could you please show the error messages you see with the latest
> master?
>

Now it fails in similar way:

./autogen.sh
[...]
Your system has the required tools.
Interfering nt/gnulib.mk from lib/gnulib.mk ...
sed: 46: "# This is the list of m ...": extra characters after \ at the 
end of c command

>
>> Maybe this can help:
>> http://lists.gnu.org/archive/html/emacs-devel/2017-01/msg00071.html
>
> That didn't work on other systems, so we cannot use it as-is.

As I wrote, on macOS sed could behave differently.. it is not a GNU sed..

>
> Thanks.
>

Re: Last commit breaks macOS build

[Eli Zaretskii]

> Cc: rcopley@gmail.com, eggert@cs.ucla.edu, emacs-devel@gnu.org
> From: Angelo Graziosi <angelo.graziosi@alice.it>
> Date: Thu, 5 Jan 2017 18:21:50 +0100
> 
> > Could you please show the error messages you see with the latest
> > master?
> >
> 
> Now it fails in similar way:
> 
> ./autogen.sh
> [...]
> Your system has the required tools.
> Interfering nt/gnulib.mk from lib/gnulib.mk ...
> sed: 46: "# This is the list of m ...": extra characters after \ at the 
> end of c command

How about now?

> >> Maybe this can help:
> >> http://lists.gnu.org/archive/html/emacs-devel/2017-01/msg00071.html
> >
> > That didn't work on other systems, so we cannot use it as-is.
> 
> As I wrote, on macOS sed could behave differently.. it is not a GNU sed..

Yes, and as I wrote, I'm trying to find a solution that will work on
all systems.

Thanks.

Re: Last commit breaks macOS build

[Andreas Schwab]

On Jan 05 2017, Eli Zaretskii <eliz@gnu.org> wrote:

> How about now?

echo is not portable wrt backslashes.

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: Last commit breaks macOS build

[Eli Zaretskii]

> From: Andreas Schwab <schwab@linux-m68k.org>
> Cc: Angelo Graziosi <angelo.graziosi@alice.it>,  rcopley@gmail.com,  eggert@cs.ucla.edu,  emacs-devel@gnu.org
> Date: Thu, 05 Jan 2017 19:18:22 +0100
> 
> On Jan 05 2017, Eli Zaretskii <eliz@gnu.org> wrote:
> 
> > How about now?
> 
> echo is not portable wrt backslashes.

Suggestions for how to fix this are welcome.  I have no access to a
macOS system, so I'm flying blind.

Re: Last commit breaks macOS build

[Andreas Schwab]

On Jan 05 2017, Eli Zaretskii <eliz@gnu.org> wrote:

>> From: Andreas Schwab <schwab@linux-m68k.org>
>> Cc: Angelo Graziosi <angelo.graziosi@alice.it>,  rcopley@gmail.com,  eggert@cs.ucla.edu,  emacs-devel@gnu.org
>> Date: Thu, 05 Jan 2017 19:18:22 +0100
>> 
>> On Jan 05 2017, Eli Zaretskii <eliz@gnu.org> wrote:
>> 
>> > How about now?
>> 
>> echo is not portable wrt backslashes.
>
> Suggestions for how to fix this are welcome.  I have no access to a
> macOS system, so I'm flying blind.

That has nothing to do with macos.

$ echo '\\'
\\
$ shopt -s xpg_echo
$ echo '\\'
\

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: Last commit breaks macOS build

[Eli Zaretskii]

> From: Andreas Schwab <schwab@linux-m68k.org>
> Cc: angelo.graziosi@alice.it,  rcopley@gmail.com,  eggert@cs.ucla.edu,  emacs-devel@gnu.org
> Date: Thu, 05 Jan 2017 19:32:05 +0100
> 
> $ echo '\\'
> \\
> $ shopt -s xpg_echo
> $ echo '\\'
> \

So you are saying that autogen.sh should do

  shopt -u xpg_echo

to countermand a possible user setting?

Strangely, whether I set xpg_echo or not, the current autogen.sh
produces the same nt/gnulib.mk, although the effect on 'echo' is as
you indicate above.  So why doesn't that option have effect here?
What am I missing?

Thanks.

Re: Last commit breaks macOS build

[Andreas Schwab]

On Jan 05 2017, Eli Zaretskii <eliz@gnu.org> wrote:

> Strangely, whether I set xpg_echo or not, the current autogen.sh
> produces the same nt/gnulib.mk, although the effect on 'echo' is as
> you indicate above.  So why doesn't that option have effect here?

Because you are relying on unspecified behaviour of the sed script.

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: Last commit breaks macOS build

[Eli Zaretskii]

> From: Andreas Schwab <schwab@linux-m68k.org>
> Cc: angelo.graziosi@alice.it,  rcopley@gmail.com,  eggert@cs.ucla.edu,  emacs-devel@gnu.org
> Date: Thu, 05 Jan 2017 19:53:17 +0100
> 
> On Jan 05 2017, Eli Zaretskii <eliz@gnu.org> wrote:
> 
> > Strangely, whether I set xpg_echo or not, the current autogen.sh
> > produces the same nt/gnulib.mk, although the effect on 'echo' is as
> > you indicate above.  So why doesn't that option have effect here?
> 
> Because you are relying on unspecified behaviour of the sed script.

Thanks.

Re: Last commit breaks macOS build

[Angelo Graziosi]

Last commit (	d1d9b376306edb77632968cbbd7dfa456f00046c) builds on macOS, 
MSW and GNU/Linux..

Thanks,
  Angelo.

Re: Last commit breaks macOS build

[Eli Zaretskii]

> Cc: rcopley@gmail.com, eggert@cs.ucla.edu, emacs-devel@gnu.org
> From: Angelo Graziosi <angelo.graziosi@alice.it>
> Date: Thu, 5 Jan 2017 23:41:56 +0100
> 
> Last commit (	d1d9b376306edb77632968cbbd7dfa456f00046c) builds on macOS, 
> MSW and GNU/Linux..

Great, thanks for testing.

Re: Last commit breaks macOS build

[Angelo Graziosi]

Il 06/01/2017 08:45, Eli Zaretskii ha scritto:
>> Cc: rcopley@gmail.com, eggert@cs.ucla.edu, emacs-devel@gnu.org
>> From: Angelo Graziosi <angelo.graziosi@alice.it>
>> Date: Thu, 5 Jan 2017 23:41:56 +0100
>>
>> Last commit (	d1d9b376306edb77632968cbbd7dfa456f00046c) builds on macOS,
>> MSW and GNU/Linux..
>
> Great, thanks for testing.
>

Yes but, maybe, someone can clarify the following...

I see that now autogen.sh prints the message:

   Inferring nt/gnulib.mk from lib/gnulib.mk ...

Now, if I understand, 'nt' should refer to the build on MSW: Why, then, 
the same message in the build on GNU/Linux and on macOS?


TIA,
  Angelo.

Re: Last commit breaks macOS build

[Paul Eggert]

On 01/06/2017 03:33 PM, Angelo Graziosi wrote:
>
> Now, if I understand, 'nt' should refer to the build on MSW: Why,
> then, the same message in the build on GNU/Linux and on macOS? 

Becuase nt/gnulib.mk is built by the developer with special tools like
'autoconf', not by the builder who merely starts with a tarball and runs
'configure; make' and should not need the special tools.

The whole thing is a bit of a mess, and the mess is largely caused by
Emacs using Automake, and Automake is needed only because of Gnulib
(because Gnulib assumes Automake for portability to older systems
lacking GNU Make). I'll look into fixing this so that Gnulib no longer
assumes Automake, so that Emacs can stop relying on Automake. Since
Emacs assumes GNU Make, it doesn't need Automake (except for the Gnulib
code, which I think I can fix).