ncurses fails to build with GCC 5.x

[Leif Leonhardy]

...
gcc-5.1 -DHAVE_CONFIG_H -I../ncurses -I../../ncurses -I. -I../include -I../../ncurses/../include -I/foo/sage-6.6-gcc-5.1.0/local/include  -D_GNU_SOURCE -DNDEBUG -O2 --param max-inline-insns-single=1200 -fPIC -c ../ncurses/lib_gen.c -o ../obj_s/lib_gen.o
In file included from ../../ncurses/curses.priv.h:321:0,
                 from ../ncurses/lib_gen.c:19:
_3081.c:837:15: error: expected ')' before 'int'
../include/curses.h:1622:56: note: in definition of macro 'mouse_trafo'
 #define mouse_trafo(y,x,to_screen) wmouse_trafo(stdscr,y,x,to_screen)
                                                        ^
Makefile:1323: recipe for target '../obj_s/lib_gen.o' failed
make[1]: *** [../obj_s/lib_gen.o] Error 1
make[1]: Leaving directory '/tmp/sage-build-tmp/ncurses-5.9.20131221/src/narrow/ncurses'
Makefile:111: recipe for target 'all' failed
make: *** [all] Error 2
Error building ncurses (narrow).

(build/pkgs/ncurses/patches/ncurses-5.9.20131221_work_around_broken_GNU_cpp_5.x.patch, self-explanatory:)

ncurses-5.9.20131221/ncurses/base/MKlib_gen.sh

@@ -62 +62 @@
 if test "${LC_CTYPE+set}"    = set; then LC_CTYPE=C;    export LC_CTYPE;    fi
 if test "${LC_COLLATE+set}"  = set; then LC_COLLATE=C;  export LC_COLLATE;  fi
 
-preprocessor="$1 -DNCURSES_INTERNALS -I../include"
+# Work around "unexpected" output of GCC 5.1.0's cpp w.r.t. #line directives
+# by simply suppressing them:
+case `$1 -dumpversion 2>/dev/null` in
+    5.[01].*)  # assume a "broken" one
+        preprocessor="$1 -P -DNCURSES_INTERNALS -I../include"
+        ;;
+    *)
+        preprocessor="$1 -DNCURSES_INTERNALS -I../include"
+esac
 AWK="$2"
 USE="$3"
 

[Leif Leonhardy]

Haven't checked whether or how upstream already deals with this (probably even by adding -P unconditionally).

Perhaps J-P wants to... :P

[Leif Leonhardy]

New commits:

​2201c9c

Add patch to let ncurses build with GCC 5.x (changed output of 'cpp')

[Leif Leonhardy]

Meanwhile found an upstream patch for the issue (not yet part of a "stable" release):

ncurses/base/MKlib_gen.sh

@@ -2 +2 @@
 #
 # MKlib_gen.sh -- generate sources from curses.h macro definitions
 #
-# ($Id: MKlib_gen.sh,v 1.46 2011/06/04 19:14:08 tom Exp $)
+# ($Id: MKlib_gen.sh,v 1.47 2014/12/06 18:56:25 tom Exp $)
 #
 ##############################################################################
-# Copyright (c) 1998-2010,2011 Free Software Foundation, Inc.                #
+# Copyright (c) 1998-2011,2014 Free Software Foundation, Inc.                #
 #                                                                            #
 # Permission is hereby granted, free of charge, to any person obtaining a    #
 # copy of this software and associated documentation files (the "Software"), #
@@ -474 +474 @@
         -e 's/gen_$//' \
         -e 's/  / /g' >>$TMP
 
+cat >$ED1 <<EOF
+s/  / /g
+s/^ //
+s/ $//
+s/P_NCURSES_BOOL/NCURSES_BOOL/g
+EOF
+
+# A patch discussed here:
+#       https://gcc.gnu.org/ml/gcc-patches/2014-06/msg02185.html
+# introduces spurious #line markers.  Work around that by ignoring the system's
+# attempt to define "bool" and using our own symbol here.
+sed -e 's/bool/P_NCURSES_BOOL/g' $TMP > $ED2
+cat $ED2 >$TMP
+
 $preprocessor $TMP 2>/dev/null \
-| sed \
-        -e 's/  / /g' \
-        -e 's/^ //' \
-        -e 's/_Bool/NCURSES_BOOL/g' \
+| sed -f $ED1 \
 | $AWK -f $AW2 \
 | sed -f $ED3 \
 | sed \

Still think my solution is a bit more robust (and easier to follow ;-) ). (Haven't actually tried upstream's fix.)

Cc'ing the spkg maintainer...

[Leif Leonhardy]

P.S.: Doesn't look like that single upstream patch alone would suffice, we'd presumably have to include more (but our package is already patched).

I'm not really keen to figure out which other parts we may need, and whether those may in turn break other things.

[François Bissey]

Well the corresponding gentoo bug is ​https://bugs.gentoo.org/show_bug.cgi?id=545114 since the same dev opened and closed it, I would say they very much have tested upstream patch.

I'll have to say I like yours better, it feels more comprehensible. But upstream will probably stick to its own fix.

Current version of ncurses in sage is a dev version from upstream (available at ​ftp://invisible-island.net/ncurses/current/ the one we have is old enough to be off the list) may be we should just upgrade to one of the latest dev version? Jean-Pierre Flori did the last upgrade may be he has an opinion on the process?

[François Bissey]

Well using ncurses-5.9-20150425 got me through. I had to remove the patch for ncurses included in sage. The first part about aclocal.m4 applied with fuzz, but the second part about configure was rejected as appearing to be a reverse patch. I take that as a news that the patch in question is no longer necessary.

I'll report back about whether or not that upgrade breaks any parts of sage.

[Leif Leonhardy]

Replying to fbissey:

Current version of ncurses in sage is a dev version from upstream (available at ​ftp://invisible-island.net/ncurses/current/ the one we have is old enough to be off the list) may be we should just upgrade to one of the latest dev version?

Well, that's the problem. Last time J-P had to add a patch again because something got borked again which was fixed in the previous version, and I cannot (and don't want to) test on various platforms, with various compilers or setups.

(The diff above is btw. from the latest devel version, but only for that file, which got fixed in an earlier version.)

[Leif Leonhardy]

I'd rather upgrade on another (less important) ticket, i.e., get some working fix into Sage quickly, without risking other issues or endless review.

[Jean-Pierre Flori]

I agree, let's update in a follow up ticket.

And I fear the XOPEN_SOURCE thing if still present is still broken on some archs. The patch surely does not apply anymore because configure was regenerated...

[Jean-Pierre Flori]

And yes we should upgrade asap as well if possible.

[Leif Leonhardy]

Thanks.

Where did you have the XOPEN_SOURCE problems?

I currently cannot test on anything but Linux x86/x86_64 (skynet and bsd.math, RIP), so I didn't even try to upgrade.

[Jean-Pierre Flori]

The XOPEN_SOURCE problem was on Solaris... See ​http://trac.sagemath.org/ticket/15268#comment:10 and #15796 for some info.

Note sure what is in latest upstream right now... See ​http://invisible-island.net/ncurses/NEWS.html#index-t20131116 and ​http://invisible-island.net/ncurses/NEWS.html#index-t20140209 and ​http://invisible-island.net/ncurses/NEWS.html#index-t20150221 for some recent changes, especially the second one. I'd say the patch is not needed anymore, but we'll have to check the source or test on Solaris.

[Julian Rüth]

This is still an issue with GCC 5.2.0. See #18977.