Re: [clisp-list] Re: Building CLISP for Windows

[Sam S. <sd...@gn...>]

> * In message <DFD875E85664D3118FA6080006277DE705216404@U8PN2>
> * On the subject of "[clisp-list] Re: Building CLISP for Windows"
> * Sent on Fri, 8 Mar 2002 11:06:10 +0100
> * Honorable "Hoehle, Joerg-Cyril" <Joe...@t-...> writes:
>
> Sam Steingold wrote:
> > see CLISP/win32*/
> > I used msvc6 with CLISP/win32msvc/makefile.msvc5 and followed the
> > instructions in CLISP/win32msvc/INSTALL
> > everything worked just dandy.
> I can confirm this, using MSVC6.
> 
> However, here are some observations of mine:
> 
> o (queens 21.0s0) still crashes.

msvs has a pretty decent debugger - maybe you could track this down?
you will have to recreate makefile:
look at Makefile.devel and add "debug" to the makemake options.

> o genclisph.exe and as a consequence clisp.h are not generated and
> there's even no makefile target for them.
> Did anyone ever build a module on MS-windows?

IIUC, you need add an interface to LoadLibrary (win32 version of dlopen)
to spvw.d:dynload_modules().
Bruno proposed using libltdl from GNU libtool.
would you like to do this?

> What is the wgenclisph.exe that makefile target clean0 may remove?

Bruno?

> I used the following to complete the makefile (somebody please add this to makemake):
> genclisph.exe : genclisph.obj
> 	$(CC) $(CFLAGS) $(CLFLAGS) genclisph.obj /Fegenclisph.exe
> clisp.h : genclisph.exe
> 	genclisph.exe > clisp.h

this was disabled on win32 together will the rest of the modules
interface.
I enabled it - re-create your makefiles and try it
(you will need some UNIX utils, like sed, for some other module
targets, and I am not sure that nmake supports things like $@ &c.)

> --------
> clisp.h is not quite correct:
> mtest.c (generated from my private mtest.lisp) includes clisp.h
> clisp.h(24) : error C2632: 'long' gefolgt von 'long' ist unzulaessig
> clisp.h(25) : error C2632: 'long' gefolgt von 'long' ist unzulaessig
> It seems like genclisp.d doesn't reproduce lispbibl.d which says:
> #elif defined(MICROSOFT)
>   typedef __int64           SLONGLONG;
>   typedef unsigned __int64  ULONGLONG;
>   #define HAVE_LONGLONG
> #endif
> and just outputs:
> typedef long long           SLONGLONG;
> for Microloft VC, which is not accepted.

fixed, thanks.

> o dirkey.lisp:
> The following functions were used but not defined:
>  LDAP::TEXT
> TEXT seems to come from package SYSTEM instead (SYSTEM::TEXT) and used
> in producing error messages (IIRC it's a marker for GETTEXT).
> 
> --------
> o One of dirkey.lisp or dirkey.c should be renamed (like affi|1,
> foreign|1, rexx|1.{d,lisp}) This will avoid dirkey.c to get
> overwritten the day dirkey.lisp will contain some FFI definition, when
> it's supposed to be generated from dirkey.d. Furthermore, makefile
> (MSVC6.0) get confused and wants to rebuild my dirkey, maybe because
> of this?? or maybe because somehow dirkey.c disappeared?

another mystery solved!!! thanks!
fixed too.

> --------
> Seen after I added MODULE(mtest) to modules.h:
> modules.i.c
> modules.h(1) : warning C4113: 'void (__cdecl *)()' weicht in der
> Parameterliste von 'void (__cdecl *)(struct module_ *)' ab
> modules.h(1) : warning C4113: 'void (__cdecl *)()' weicht in der Parameterliste
> von 'void (__cdecl *)(struct module_ *)' ab
> "x deviates in the parameter list from y."
> Looks like something was missing or not expanded. That doesn't look like a good definition.
> 
> I also have 5KB worth of compiler warnings from varying C files (in German, sorry). Where should I send that?

Please fix them and send in a patch instead! :-)

-- 
Sam Steingold (http://www.podval.org/~sds) running RedHat7.2 GNU/Linux
Keep Jerusalem united! <http://www.onejerusalem.org/Petition.asp>
Read, think and remember! <http://www.iris.org.il> <http://www.memri.org/>
Bus error -- driver executed.