readline: link directly against ncurses

---
note: not really tested since external toolchain support is totally broken

 package/readline/readline-6.1-curses-link.patch |   16 ++++++++++++++++
 1 files changed, 16 insertions(+), 0 deletions(-)
 create mode 100644 package/readline/readline-6.1-curses-link.patch

new file mode 100644
index 0000000..56b4770
@@ -0,0 +1,16 @@
+link readline directly to ncurses since it needs symbols from it
+
+upstream readline does this on purpose (no direct linking), but
+it doesn't make much sense in a Linux world
+
+--- a/support/shobj-conf
++++ b/support/shobj-conf
+@@ -42,7 +42,7 @@
+ SHOBJ_LIBS=
+ 
+ SHLIB_XLDFLAGS=
+-SHLIB_LIBS=
++SHLIB_LIBS=-lncurses
+ 
+ SHLIB_DOT='.'
+ SHLIB_LIBPREF='lib'

1.7.4.1

Hi Mike

I can't thank you enough....if it wasn't for yourself and Joakim I would still have no idea.

I manually tried your "sed" suggestion and can confirm it worked:

[root at dlxarm01 buildroot]# readelf -d ./output/target/usr/lib/libreadline.so.6.1 | grep "]"

 0x00000001 (NEEDED)                     Shared library: [libncurses.so.5]
 0x00000001 (NEEDED)                     Shared library: [libc.so.0]
 0x0000000e (SONAME)                     Library soname: [libreadline.so.6]
 0x0000000f (RPATH)                      Library rpath: [/usr/lib]

[root at dlxarm01 buildroot]#

And now when I run python:
[root at vx-200 ~]# python
Python 2.7.1 (r271:86832, Mar 23 2011, 23:27:09)
[GCC 4.5.2] on linux2
Type "help", "copyright", "credits" or "license" for more information.

[root at vx-200 ~]#

:))))))))))))))))))))))))))))))))))))))))))))))))))))))

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.busybox.net/pipermail/buildroot/attachments/20110323/ea136681/attachment.html>

Hello,

Could you elaborate ?

Thomas

Thomas Petazzoni, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com

i'm pretty sure i pointed this out around the time it broke. if you dont specify the full path, the stupid makefiles prepend garbage.

BR2_TOOLCHAIN_EXTERNAL=y BR2_TOOLCHAIN_EXTERNAL_CUSTOM=y BR2_TOOLCHAIN_EXTERNAL_PATH="" BR2_TOOLCHAIN_EXTERNAL_CUSTOM_PREFIX="bfin-uclinux" BR2_TOOLCHAIN_EXTERNAL_PREFIX="bfin-uclinux"

$ make ncurses
....
checking for bfin-linux-gcc... /bin/bfin-uclinux-gcc
--sysroot=/usr/local/src/buildroot/output/host/usr/bfin-unknown-linux-uclibc/sysroot
-mike

Mike> i'm pretty sure i pointed this out around the time it broke. if you Mike> dont specify the full path, the stupid makefiles prepend garbage.

Mike> BR2_TOOLCHAIN_EXTERNAL=y
Mike> BR2_TOOLCHAIN_EXTERNAL_CUSTOM=y
Mike> BR2_TOOLCHAIN_EXTERNAL_PATH=""
Mike> BR2_TOOLCHAIN_EXTERNAL_CUSTOM_PREFIX="bfin-uclinux" Mike> BR2_TOOLCHAIN_EXTERNAL_PREFIX="bfin-uclinux"

Ahh, that got reported some time ago:

http://lists.busybox.net/pipermail/buildroot/2011-March/041891.html

I'll fix it now.

Bye, Peter Korsgaard

Come on Mike, just for this reason you call the external toolchain support "totally broken" ? It's fixed by a two-liner patch that I acked some time ago:
http://lists.busybox.net/pipermail/buildroot/2011-March/041901.html.

Regards,

Thomas

Thomas Petazzoni, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com

the size of the patch doesnt make it any less broken. and since it wasnt merged until just now, i couldnt test anything in mainline. -mike

Mike> Signed-off-by: Mike Frysinger <[email protected]>

Committed, thanks.

Bye, Peter Korsgaard