Finch and console ui

[Joshua Blanton]

Bill Fassler wrote:
> Although I am attempting to build ELF flat static, I am still
> having success up until panel symbols.  According to my config.log
> the autoconf utilities find glib, and then they cannot find the
> wide versions of ncurses or panel, but then they fall back on the
> narrow versions (as the script infers that they should).

Yeah, the fallback should work.  At least, I've heard people report
that it does - I've never tried to build against ncurses/panel, and
certainly not statically.

> **************************************
> But does find panel
> *************************************
> configure:28170: checking for update_panels in -lpanel
> configure:28205: bfin-uclinux-gcc -o conftest -D__linux__ -DNOMMU -I/home/vocal/project/blackfin-svn-branch/elf_flat/uClinux-dist/staging/usr/include -D__lin
> ux__ -DNOMMU -I/home/vocal/project/blackfin-svn-branch/elf_flat/uClinux-dist/staging/usr/include -lpthread -L/home/vocal/project/blackfin-svn-branch/elf_flat
> /uClinux-dist/staging/usr/lib conftest.c -lpanel  -lnsl -lresolv  >&5
> /home/vocal/project/blackfin-svn-branch/elf_flat/uClinux-dist/staging/usr/lib/libpanel.a(p_update.o): In function `_update_panels':
> ****************************************
> But fails here... I am not positive that this is the last hurdle
> to a console, but I hope so.  I just don't understand this one yet.
> Since I am building ELF flat static, I wonder it the pidgin/finch
> link commands are in the right order for dependencies.  I would venture
> to guess that static builds of Finch are not often (if ever) tested.
> In any event I solicit anyones opinion here..
> ****************************************************************
> ../../panel/p_update.c:(.text+0x8): undefined reference to `__nc_panelhook'
> ../../panel/p_update.c:(.text+0x8c): undefined reference to `_is_linetouched'
> ../../panel/p_update.c:(.text+0xf0): undefined reference to `__nc_panelhook'
> ../../panel/p_update.c:(.text+0x102): undefined reference to `_wnoutrefresh'
> collect2: ld returned 1 exit status
> configure:28211: $? = 1
> configure: failed program was:
> 
> Any ideas?  Is it possible that the order of dependent libraries
> are not correct since almost everybody uses shared?

nm on my machine says that those functions are exported by
libncurses.a, so the test compile should probably use -lncurses as
well; I'm not sure why it doesn't.  Perhaps because, when building
with dynamic libraries, libpanel/libpanelw list dependencies on
libncurses/libncursesw?  I don't remember exactly how much
dependency resolution gcc will do for you in that case...  From the
compile line, though, I'd infer "enough" ;-)

I assume, though I don't know for sure, that autoconf provides the
test for panel/ncurses.  If that's the case, their script is
busted.  If it's included with pidgin, that script should be fixed.

--jtb
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
URL: <http://pidgin.im/pipermail/devel/attachments/20070905/da767bb6/attachment.sig>