On Thu, 2014-06-26 at 17:53 +0200, Vincent Bernat wrote: > ❦ 26 juin 2014 10:16 -0500, Dan Williams <[email protected]> : > > > What does 'ldd' on your libreadline.so return? Mine has: > > > > linux-vdso.so.1 => (0x00007fff191fe000) > > libtinfo.so.5 => /lib64/libtinfo.so.5 (0x0000003473400000) > > libc.so.6 => /lib64/libc.so.6 (0x000000345ac00000) > > /lib64/ld-linux-x86-64.so.2 (0x000000345a800000) > > I have a similar output: > > linux-vdso.so.1 (0x00007fff2c9fe000) > libtinfo.so.5 => /lib/x86_64-linux-gnu/libtinfo.so.5 > (0x00007f98d71e2000) > libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007f98d6e39000) > /lib64/ld-linux-x86-64.so.2 (0x00007f98d768a000) > > > so at least the dynamic linker knows that libtinfo (provided by ncurses) > > is required here, and should load that in for the configure test > > program. > > The dynamic linker knows but the linker used to build the executables > may not resolve symbols from one object with the dependencies of another > object. Only symbols of explicitely linked libraries may be used. > > GNU linker will try to resolve dependencies recursively but more modern > linkers like the Gold linker doesn't. GNU linker may be told not to do > that with -Wl,--no-add-needed. Reading the documentation, I see that > --no-add-needed is now --no-copy-dt-needed-entries and I am a bit lazy > to understand the explanation. Also read the --as-needed option: > https://sourceware.org/binutils/docs/ld/Options.html > > Some distribution enforce those flags.
Yeah, I forgot about gold. I guess this is why pkgconfig is great, since then you don't have to do a crapload of guesswork yourself (or in the m4). So, I suppose we add the readline m4 thing, but unfortunately we need to hack it up ourselves, because it doesn't define READLINE_LIBS even though it did all the work to find them... Dan _______________________________________________ networkmanager-list mailing list [email protected] https://mail.gnome.org/mailman/listinfo/networkmanager-list
