Re: DarwinPorts 1.3.1

On Jul 31, 2006, at 10:00 PM, Paul Guyot wrote:
>
> Le 1 août 06 à 09:52, Anthony Agelastos a écrit :
>
>> On Jul 31, 2006, at 7:56 PM, Juan Manuel Palacios wrote:
>>>
>>> 	Readline compilation is failing for you:
>> This is the synopsis I have received from the three people who  
>> have replied thus far. So, I took jk's suggestion and uninstalled  
>> readline (and gnuplot) and then re-did the sudo port -d selfupdate  
>> and I still have, more or less, the same problem. I have attached  
>> the latest version of the build for an environment where I have  
>> deinstalled readline.
>>>
>>> 	The symbols that were previously determined to exist in your  
>>> system libraries weren't really found. Have you performed any  
>>> type of surgery on your system? Particularly with the readline  
>>> libraries?
>> I have not performed any surgery on my system. I have Gerben  
>> Wierda's MacTeX and Gaurav Khanna's Octave installed in /usr/ 
>> local ; other than that, this machine is Apple-standard. Any  
>> ideas? I am aware that I can probably just download the binary  
>> updater and go from there. I mention this problem with selfupdate  
>> to assist the developers in finding any potential software bugs. I  
>> hope this helps. Thank you for the quick replies, and thank you to  
>> the developers who have made DarwinPorts a great tool.
>
> Anthony,
>
> You have a copy of readline in /usr/local.
> This copy is a recent version unlike the one in /usr/lib and it  
> defines symbols with an rl_ prefix:
This readline is from an installation of Octave pre-packaged from  
Gaurav Khanna at hpc.sf.net for Intel Macs.

>
>> checking whether rl_username_completion_function is declared... yes
>> checking whether rl_filename_completion_function is declared... yes
>> checking whether rl_completion_matches is declared... yes
>> checking whether username_completion_function is declared... no
>> checking whether filename_completion_function is declared... no
>> checking whether completion_matches is declared... no
>
> But here, ld tries to link with the copy in /usr/lib which doesn't  
> have such symbols:
>
>> cc -dynamiclib -L/usr/lib -lcurl -lssl -lcrypto -lz Pextlib.o  
>> strsed.o fgetln.o md5cmd.o setmode.o xinstall.o find.o  
>> strcasecmp.o vercomp.o filemap.o sha1cmd.o compat.o curl.o  
>> rmd160cmd.o readline.o uid.o -o Pextlib.dylib -L/System/Library/ 
>> Frameworks/Tcl.framework/Versions/8.4 -ltclstub8.4 -L/usr/lib - 
>> lcurl -lssl -lcrypto -lz -lreadline  -lcrypto
>> ld: warning prebinding disabled because of undefined symbols
>> ld: Undefined symbols:
>> _rl_completion_matches
>> _rl_filename_completion_function
>> _rl_username_completion_function
>
> The only good solution to this would be to add /usr/local/lib in  
> the library search path when building base/ but it's contrary to  
> the tradition of darwinports which says that /usr/local is evil.  
> Preventing gcc to use /usr/local/include would cause way too much  
> harms by preventing the software to be compiled on non-standard  
> installs.
 From other posts, I have gathered that it seems that during one part  
of the install process, it is using readline in /usr/lib and in the  
other, it is using the one in /usr/local/lib. Why is it mixing? How  
does my system even know about the readline in /usr/local/lib? It was  
installed via tar, so all that it did was copy files in place. If it  
is because I have /usr/local/bin in my PATH, then I should note that / 
usr/local/bin is the second-to-last entry in my PATH with /usr/local/ 
teTeX/bin/... the last one and that /usr/bin is earlier in my PATH. I  
would expect the system to default the standard readline. What pieces  
am I missing? Thank you for your assistance with this.

And, btw, to add the the library search path, is it

export LD_LIBRARY=$LD_LIBRARY:/usr/local/lib

? I don't recall if it is LD_LIBRARY or something else.

>
> Paul
> -- 
> Ministre ultraplénipotentiaire en disponibilité.
> Mobile. Sans baignoire fixe.
> http://www.kallisys.com/
> http://www-poleia.lip6.fr/~guyot/
>