Build also wide version of ncurses

[vbraun]

Curses optionally supports wchar versions, the library will be called libncursesw.so. Python wants the wide version for its curses module, so we should build it as well.

Updated spkg: ​http://boxen.math.washington.edu/home/vbraun/spkg/ncurses-5.9.p1.spkg

Apply trac_15080.patch​

[vbraun]

It seems that Python will fall back to narrow curses if necessary. But if the distribution provides a wide curses then it will be preferred over the narrow one. On my Fedora 19 install, this leads to (with the old ncurses spkg):

building '_curses' extension
gcc -pthread -fPIC -fno-strict-aliasing -g -O2 -DNDEBUG -g -fwrapv -O3 -Wall -I/home/vbraun/opt/sage-5.12.beta3/local/include -I. -IInclude -I./Include -I/usr/local/include -I/home/vbraun/opt/sage-5.12.beta3/spkg/build/python-2.7.5.p1/src/Include -I/home/vbraun/opt/sage-5.12.beta3/spkg/build/python-2.7.5.p1/src -c /home/vbraun/opt/sage-5.12.beta3/spkg/build/python-2.7.5.p1/src/Modules/_cursesmodule.c -o build/temp.linux-x86_64-2.7/home/vbraun/opt/sage-5.12.beta3/spkg/build/python-2.7.5.p1/src/Modules/_cursesmodule.o
gcc -pthread -shared -L/home/vbraun/opt/sage-5.12.beta3/local/lib -L/home/vbraun/opt/sage-5.12.beta3/local/lib build/temp.linux-x86_64-2.7/home/vbraun/opt/sage-5.12.beta3/spkg/build/python-2.7.5.p1/src/Modules/_cursesmodule.o -L/home/vbraun/opt/sage-5.12.beta3/local/lib -L/usr/local/lib -L. -lncursesw -lpython2.7 -o build/lib.linux-x86_64-2.7/_curses.so
*** WARNING: renaming "_curses" since importing it failed: /lib64/libncursesw.so.5: undefined symbol: _nc_putchar

and it will succeed with the new spkg.

[jpflori]

Just a few random remarks: why use ? rather than plain [ ] for testing? Could you mention the ticket number in SPKG.txt? I know it may seem overkill but it always nice to have all info available at once...

[jpflori]

Otherwise looks fine. I'll still have to work on top of that to solve the Solaris problems I reported at #15268 (disable ada and use lmonade fix for CPPFLAGS vs CFLAGS).

[vbraun]

Replying to jpflori:

Could you mention the ticket number in SPKG.txt?

IMHO we should delete the duplicate changelog keeping as soon as we have a working git version. Its just a stupid waste of developer time.

[vbraun]

Updated spkg at same place

[jdemeyer]

$SAGE_LOCAL/bin/.hgignore needs to be updated:

$ hg status
? ncursesw5-config

[jpflori]

update hgignore

[jpflori]

Apply trac_15080.patch​

[jdemeyer]

.hgignore is good.

[jdemeyer]

This fails to build on Solaris SPARC due to invalid CFLAGS:

/home/buildbot/build/sage/mark-1/mark_full/build/sage-5.13.beta3/local/bin/g++ -I../c++ -I../include -I../../c++ -DHAVE_CONFIG_H   -D__EXTENSIONS__ -D_XOPEN_SOURCE_EXTENDED -D_FILE_OFFSET_BITS=64  -DNDEBUG -I. -I../include -I../../c++/../include -I/home/buildbot/build/sage/mark-1/mark_full/build/sage-5.13.beta3/local/include  -fPIC -c ../../c++/cursesf.cc -o ../obj_s/cursesf.o
/home/buildbot/build/sage/mark-1/mark_full/build/sage-5.13.beta3/local/bin/g++ -I../c++ -I../include -I../../c++ -DHAVE_CONFIG_H   -D__EXTENSIONS__ -D_XOPEN_SOURCE_EXTENDED -D_FILE_OFFSET_BITS=64  -DNDEBUG -I. -I../include -I../../c++/../include -I/home/buildbot/build/sage/mark-1/mark_full/build/sage-5.13.beta3/local/include  -fPIC -c ../../c++/cursesm.cc -o ../obj_s/cursesm.o
In file included from /home/buildbot/build/sage/mark-1/mark_full/build/sage-5.13.beta3/local/lib/gcc/sparc-sun-solaris2.10/4.7.3/include-fixed/iso/stdlib_iso.h:39:0,
                 from /usr/include/stdlib.h:18,
                 from ../../c++/internal.h:53,
                 from ../../c++/cursesf.cc:34:
/home/buildbot/build/sage/mark-1/mark_full/build/sage-5.13.beta3/local/lib/gcc/sparc-sun-solaris2.10/4.7.3/include-fixed/sys/feature_tests.h:341:2: error: #error "Compiler or options invalid for pre-UNIX 03 X/Open applications 	and pre-2001 POSIX applications"

[jpflori]

Hum, sure, I forgot about #15268 which deals with my ada issue and the xopen_source_extended one. I'll provide an updated spkg there.

[jpflori]

Spkg uploaded at #15268, based on the one here.

[vbraun]

So let's close this ticket, further work at #15268.

[jpflori]

Volker: On your Red Hat, with what got readline linked? I'm currently looking into Python's install scripts and it seems fishy Python ended up wanting to uses ncursesw on top of ncurses unless readline was linked to it already.

[jpflori]

Or not if readline was only linked to tinfo or even to nothing as seems to be the case on Red Hat...

[jpflori]

Ok think I got it, more work for me. As a consequence the ticket here should not really be necessary anymore but won't hurt...

[vbraun]

For the record, readline is just linking to the standard ncurses on Fedora 19. It is only the Python curses module that links in ncursesw:

(sage-sh) vbraun@localhost:sage-5.12$ ldd local/lib/python2.7/lib-dynload/_curses.so
	linux-vdso.so.1 =>  (0x00007fff433a7000)
	libncursesw.so.5 => /home/vbraun/Code/sage/local/lib/libncursesw.so.5 (0x00007f5842950000)
	libpython2.7.so.1.0 => /home/vbraun/Code/sage/local/lib/libpython2.7.so.1.0 (0x00007f5842547000)
	libpthread.so.0 => /lib64/libpthread.so.0 (0x00007f58422fe000)
	libc.so.6 => /lib64/libc.so.6 (0x00007f5841f3d000)
	libtinfow.so.5 => /home/vbraun/Code/sage/local/lib/libtinfow.so.5 (0x00007f5841d09000)
	libdl.so.2 => /lib64/libdl.so.2 (0x00007f5841b05000)
	libutil.so.1 => /lib64/libutil.so.1 (0x00007f5841902000)
	libm.so.6 => /lib64/libm.so.6 (0x00007f58415ff000)
	/lib64/ld-linux-x86-64.so.2 (0x0000003895000000)
(sage-sh) vbraun@localhost:sage-5.12$ ldd local/lib/libreadline.so
	linux-vdso.so.1 =>  (0x00007fff75dd2000)
	libtinfo.so.5 => /home/vbraun/opt/sage-5.12/local/lib/libtinfo.so.5 (0x00007faefecae000)
	libc.so.6 => /lib64/libc.so.6 (0x00007faefe8bf000)
	/lib64/ld-linux-x86-64.so.2 (0x0000003895000000)
(sage-sh) vbraun@localhost:sage-5.12$ ldd local/lib/libtinfo.so
	linux-vdso.so.1 =>  (0x00007fff1d969000)
	libc.so.6 => /lib64/libc.so.6 (0x00007f9872a87000)
	/lib64/ld-linux-x86-64.so.2 (0x0000003895000000)

[jpflori]

Can you ldd /lib64/libncursew... please?

[jpflori]

For future ref, and so that i don't waste 5 minutes each time looking for it: ​http://bugs.python.org/issue7384.

[vbraun]

$ cat /lib64/libncursesw.so
INPUT(libncursesw.so.5 -ltinfo)

$ ldd /lib64/libncursesw.so.5
	linux-vdso.so.1 =>  (0x00007fff817fe000)
	libc.so.6 => /lib64/libc.so.6 (0x0000003895800000)
	libdl.so.2 => /lib64/libdl.so.2 (0x0000003895c00000)
	libtinfo.so.5 => /lib64/libtinfo.so.5 (0x00000038adc00000)
	/lib64/ld-linux-x86-64.so.2 (0x0000003895000000)