[LiDIA] 2.1Pre7 on Mac OS X

[Justin Walker]

Update: the 'weak link' warning actually refers to a library that was  
not built on Mac OS X 10.2 (gmp), and is likely harmless.  Weak  
linking, FWIW, refers to the ability of the linker to permit the  
linking of an object against a library, and not complain if some  
'external' symbols don't resolve (aren't present).  The app will blow  
up, of course, if it tries to use such symbols, but their absence won't  
inhibit running otherwise.  Support for this is new in 10.2.

On Sunday, Feb 2, 2003, at 14:51 US/Pacific, Justin Walker wrote:

> Christian/Christoph,
>
> It works!
>
> Here's what I've found:
>  - if I do nothing other than build the libtool package (version  
> 1.4.3) and
>    place the resulting 'libtool' in the LiDIA2.1pre7 directory, *and*  
> make
>    sure that "." is in my PATH, then the build completes successfully
>  - the examples build and (quick test only) execute correctly (with a  
> proviso
>    I'll get to below)
>  - if I compile a test program of my own, it works (see proviso).
>
> The proviso is that I get a warning during the link phase, both for  
> the examples and my own tests, as follows:
>
> ld: warning dynamic shared library: /usr/local/lib/libgmp.dylib not  
> made a weak library in output with MACOSX_DEPLOYMENT_TARGET  
> environment variable set to: 10.1
>
> This has been discussed on Mac OS X/Darwin mailing lists, but I've  
> payed no attention, so I'm not sure what the issue is, whether it's  
> important, nor how to "fix" it.  I will track this one down in the  
> next several days and report back.
>
> It is a warning, and doesn't affect (as far as I can tell) the  
> libraries or the code that links with them.
>
> A couple of other points:
>  - I have yet to try out this new library with 'hecke'; I'll do
>    that and report as well
>  - I can't say for other systems, but on Mac OS X/Darwin, if you
>    have two versions of the same library (libfoo.a and libfoo.dylib),
>    there is no way to tell the build system to use the static version.
>    The only way I've found to deal with this is to rename one.  FWIW,  
> I needed
>    this ability to build a version of Prof. Cremona's 'mwrank' suite  
> totally
>    statically (so that one could download the programs from his site  
> and run
>    them, without having to worry about how to get the right versions  
> of the
>    right libraries).  Is this a general issue?
>
> Thanks for your suggestions.  They made the difference.
>
> Regards,
>
> Justin
>
> On Saturday, Feb 1, 2003, at 11:40 US/Pacific, Christian Cornelssen  
> wrote:
>
>> On Sat, 1 Feb 2003, Justin Walker wrote:
>>
>>> Thanks for the quick reply.  I run 'aclocal' without error, but
>>> 'autoconf' (2.54) gives
> [snip]
>>
>> Could that be one of those problems that Christoph has experienced?
>> There are nasty problems with autotool installations.  Try to install
>> autoconf, libtool, and automake with the same --prefix, and in that  
>> order
>> (or in another, perhaps you need a second round!), and verify that
>> these tools detect each other correctly (during configuration) and  
>> don't
>> occasionally use macro directories from previous/system-wide versions.
> [snip]
>
> --
> Justin C. Walker, Curmudgeon-At-Large  *
> Institute for General Semantics        | Some people have a got a  
> mental
>                                        |  horizon of radius zero, and  
> call
>                                        |  it their point of view
>                                        |     -- David Hilbert
> *-------------------------------------- 
> *-------------------------------*
>
> _______________________________________________
> LiDIA mailing list
> LiDIA at cdc.informatik.tu-darmstadt.de
> http://www.cdc.informatik.tu-darmstadt.de/mailman/listinfo/lidia
>
--
Justin C. Walker, Curmudgeon-At-Large  *
Institute for General Semantics        |   If you're not confused,
                                        |   You're not paying attention
*--------------------------------------*-------------------------------*