[clisp-list] modules and FFI dependency (was: Compiling clisp-2.30: No rule to make target `clisp-module')

[Pascal B. <pj...@in...>]

Hoehle, Joerg-Cyril writes:
> Hi,
> 
> For completeness, the below message from "./configure" is wrong. I
> consider it more to be describing a limitation of configure's
> --with-module extension or the clisp-link sh script.
> [...] 
> The quoted text below should say no more than:
> 
> The modules/ subdirectory contains external CLISP modules. On UNIX,
> modules can be linked to CLISP using the clisp-link shell script. This
> is what configure --with-module=X will prepare. Modules using
> DEF-CALL-IN/OUT forms depend on the FFI.
> 
> It is a limitation of the configure [and clisp-link?] script that
> using --with-module=X turns on --with-dynamic-ffi [but who would
> build CLISP without it except for testing?]
> 
> 
> Pascal Bourguignon a écrit:
> >Or alternatively,  have a  little note somewhere  about the  fact that
> >modules should be found deeply.  Perhaps in configure, since I infered
> >(cd modules ; for f in * ; ... ) from:
> >
> >         --with-module=MODULE  build with add-on MODULE
> >                               this requires dynamic FFI and will turn it on
> >                               if the FFI checks pass
> >-->
> >         --with-module=MODULE  build with add-on MODULE, name of directory,
> >                               or one of the mutually exclusive
> >                               directory/subdirectory, in ./modules/
> >                               This requires dynamic FFI and will turn it on
> >                               if the FFI checks pass
> 
> What's mutually exclusive? Postgres 6.x and 6.y? Why? Incompatible
> link libraries?

Perhaps, the fact that both modules export the same SQL package?

Or   the  same   LINUX   and  LINUX-AUX   packages   defined  by   the
bindings/linuxlibc* modules?

Granted, one perhaps  could compile all of them at  once, but we could
not build a system  with all of them.

The point is  that a naive user cannot  blindly configure and compile.
Perhaps it should be added in  the link.sh scripts tests to see if the
module can  be compiled for the  current target (or this  could be the
job  of  configure),  so  amigaos  would not  include  itself  to  the
compilation when  the amigaos libraries  are not available, and  so on
for the  linuxlibcs.  Thus, the  right bindings for the  target system
could be selected automatically.


> Thanks for trying to compile as much modules as possible and reporting.

Thanks for providing us with these modules! 
 
> Regards,
> 	Jörg Höhle.



-- 
__Pascal_Bourguignon__                   http://www.informatimago.com/
----------------------------------------------------------------------
There is no worse tyranny than to force a man to pay for what he does not
want merely because you think it would be good for him.  -- Robert Heinlein