I am trying to install gnuplot from source on a MacBookPro running Leopard (Mac OS 10.5). I have tried to compile the official sources for gnuplot 4.2.0, 4.2.2 and the latest version from the CVS repository (as of 8Feb2008). I do:
./configure make sudo make install
and at the make stage I always (for all the above versions) get the following errors:
Undefined symbols: "_rl_forced_update_display", referenced from: _restore_prompt in command.o "_rl_ding", referenced from: _alert in mouse.o "_history_list", referenced from: _write_history_list in history.o "_rl_complete_with_tilde_expansion", referenced from: _rl_complete_with_tilde_expansion$non_lazy_ptr in plot.o "_rl_reset_after_signal", referenced from: _main in plot.o ld: symbol(s) not found collect2: ld returned 1 exit status make[3]: *** [gnuplot] Error 1 make[2]: *** [all-recursive] Error 1 make[1]: *** [all-recursive] Error 1 make: *** [all] Error 2
Any help or suggestions for diagnosing would be much appreciated.
In the meantime, I have been using the gnuplot4.2.2 binary distributed inside the distribution by the Octave project. This can be found at:
You download the octave-3.0.0-i386.dmg (for intel macs, there is also a PPC version)
mount it (without installing), and inside of the "extras" folders there is a file called gnuplot-4.2.2-i386.dmg which can be installed on its own.
Although this binary works well, I would really appreciate some help in getting the "right" thing to work so that I can upgrade and test the newer versions.
>I am trying to install gnuplot from source on a MacBookPro running >Leopard (Mac OS 10.5). I have tried to compile the official sources >for gnuplot 4.2.0, 4.2.2 and the latest version from the CVS >repository (as of 8Feb2008). I do:
>./configure >make >sudo make install
>and at the make stage I always (for all the above versions) get the >following errors:
>Undefined symbols: > "_rl_forced_update_display", referenced from: > _restore_prompt in command.o > "_rl_ding", referenced from: > _alert in mouse.o > "_history_list", referenced from: > _write_history_list in history.o > "_rl_complete_with_tilde_expansion", referenced from: > _rl_complete_with_tilde_expansion$non_lazy_ptr in plot.o > "_rl_reset_after_signal", referenced from: > _main in plot.o >ld: symbol(s) not found
Please see tracker item #102055 on the SourceForge site.
In a nutshell, this is a problem with OSX itself. Apple chose to ship Leopard with an emulated libreadline package. But the emulation is incomplete, and programs that link against it expecting a "real" libreadline fail with erros like the ones you quote. I suggest complaining to Apple. They should hear from their users that their choice of libraries was not good.
Meanwhile, your options are either
(1) delete the broken libreadline that shipped with Leopard and replace it with the real gnu libreadline
(2) tell gnuplot not to use libreadline at all ./configure --with-readline=builtin
Thanks! The change in libreadline option solved the compilations issues (I'll complain to Apple!).
However, a quick test of the install reveals that on launch gnuplot has the terminal set to "unkown" and attempts to set it to a useful interactive terminal such as "aqua" or "X" are not sucessful.
This is what my session looks like:
### Terminal type set to 'unknown' gnuplot> set terminal X ^ unknown or ambiguous terminal type; type just 'set terminal' for a list ###
I typed:
set term
for info on the available terminals and neither "aqua" nor "x" appear on the list. In contrast, the same listing from the binary release from the Octave version gives "aqua", "x11" and "xlib" as possibilities. Any idea of how to get back the missing terminals? Or is this issue connected to the libreadline misbehaviour?
Thanks a lot for the help - antia
On Feb 8, 12:02 pm, merr...@u.washington.edu (Ethan Merritt) wrote:
> >I am trying to install gnuplot from source on a MacBookPro running > >Leopard (Mac OS 10.5). I have tried to compile the official sources > >for gnuplot 4.2.0, 4.2.2 and the latest version from the CVS > >repository (as of 8Feb2008). I do:
> >./configure > >make > >sudo make install
> >and at the make stage I always (for all the above versions) get the > >following errors:
> >Undefined symbols: > > "_rl_forced_update_display", referenced from: > > _restore_prompt in command.o > > "_rl_ding", referenced from: > > _alert in mouse.o > > "_history_list", referenced from: > > _write_history_list in history.o > > "_rl_complete_with_tilde_expansion", referenced from: > > _rl_complete_with_tilde_expansion$non_lazy_ptr in plot.o > > "_rl_reset_after_signal", referenced from: > > _main in plot.o > >ld: symbol(s) not found
> Please see tracker item #102055 on the SourceForge site.
> In a nutshell, this is a problem with OSX itself. > Apple chose to ship Leopard with an emulated libreadline package. > But the emulation is incomplete, and programs that link against it expecting > a "real" libreadline fail with erros like the ones you quote. > I suggest complaining to Apple. They should hear from their users that > their choice of libraries was not good.
> Meanwhile, your options are either
> (1) delete the broken libreadline that shipped with Leopard > and replace it with the real gnu libreadline
> (2) tell gnuplot not to use libreadline at all > ./configure --with-readline=builtin
>However, a quick test of the install reveals that on launch gnuplot >has the terminal set to "unkown" and attempts to set it to a useful >interactive terminal such as "aqua" or "X" are not sucessful.
I have to assume that the ./configure could not find support libraries or appropriate header files for aquaterm and x11. But I know this works in general, so I can't suggest anything other than making sure you really do have the relevant Apple developers' kits installed.
You can get more specific information on what it failed to find by looking in the file config.log, which is a record of what the ./configure script did.
>### >Terminal type set to 'unknown' >gnuplot> set terminal X > ^ > unknown or ambiguous terminal type; type just 'set terminal' >for a list >###
>I typed:
>set term
>for info on the available terminals and neither "aqua" nor "x" appear >on the list. In contrast, the same listing from the binary release >from the Octave version gives "aqua", "x11" and "xlib" as >possibilities. Any idea of how to get back the missing terminals? Or >is this issue connected to the libreadline misbehaviour?
>Thanks a lot for the help - antia
>On Feb 8, 12:02 pm, merr...@u.washington.edu (Ethan Merritt) wrote: >> In article <259f4a58-9565-46a2-99e2-ca6f628ce...@q21g2000hsa.googlegroups.com>,
>> >I am trying to install gnuplot from source on a MacBookPro running >> >Leopard (Mac OS 10.5). I have tried to compile the official sources >> >for gnuplot 4.2.0, 4.2.2 and the latest version from the CVS >> >repository (as of 8Feb2008). I do:
>> >./configure >> >make >> >sudo make install
>> >and at the make stage I always (for all the above versions) get the >> >following errors:
>> >Undefined symbols: >> > "_rl_forced_update_display", referenced from: >> > _restore_prompt in command.o >> > "_rl_ding", referenced from: >> > _alert in mouse.o >> > "_history_list", referenced from: >> > _write_history_list in history.o >> > "_rl_complete_with_tilde_expansion", referenced from: >> > _rl_complete_with_tilde_expansion$non_lazy_ptr in plot.o >> > "_rl_reset_after_signal", referenced from: >> > _main in plot.o >> >ld: symbol(s) not found
>> Please see tracker item #102055 on the SourceForge site.
>> In a nutshell, this is a problem with OSX itself. >> Apple chose to ship Leopard with an emulated libreadline package. >> But the emulation is incomplete, and programs that link against it expecting >> a "real" libreadline fail with erros like the ones you quote. >> I suggest complaining to Apple. They should hear from their users that >> their choice of libraries was not good.
>> Meanwhile, your options are either
>> (1) delete the broken libreadline that shipped with Leopard >> and replace it with the real gnu libreadline
>> (2) tell gnuplot not to use libreadline at all >> ./configure --with-readline=builtin
Thank you both for the pointers to the binary and source of the code compilation problems on Leopard respectively. I spent the last hour fiddling with this issue (http://twitter.com/codesurgeon/statuses/ 782232753).
