On 9/29/07, Bram Moolenaar <[EMAIL PROTECTED]> wrote:
>
> Guopeng Wen -
>
> > On 8/11/07, Bram Moolenaar <[EMAIL PROTECTED]> wrote:
> > >
> > > Chris Sutcliffe wrote:
> > >
> > > > > Thanks for taking the effort to verify what works.
> > > >
> > > > No problem at all, I like to contribute where I can!
> > > >
> > > > > I'm a bit confused though.  In the last part of the patch
> > > > > __IID_DEFINED__ if it wasn't defined yet.  Why the extra complicated 
> > > > > #if
> > > > > to define it above this?  Was it defined before but to something else?
> > > >
> > > > __IID_DEFINED__ was originally passed by the MinGW makefile because
> > > > the w32api didn't explicitly define it, despite the fact that w32api
> > > > did have a definition for IID (used in the OLE stuff).  As a result,
> > > > the compile would error out because of the duplicate IID definition.
> > > > With the latest w32api (3.10), __IID_DEFINED__ is now properly defined
> > > > along with IID itself.  As a result, when the Makefile ran, it caused
> > > > a duplicate definition of __IID_DEFINED__.
> > > >
> > > > Basically what I've done is that in iid_ole.c (where IID is defined,
> > > > but guarded by __IID_DEFINED__), I've added an extra check to
> > > > essentially accomplish what the Makefile was previously doing with the
> > > > -D__IID_DEFINED__ option for previous versions of the w32api.
> > >
> > > Ah, I missed that after the #ifdef __IID_DEFINED__ the structure was
> > > declared.  I'll include the patch now, thanks.
> > >
> >
> > I found the same problem with Make_cyg.mak on the latest cygwin when
> > I try to build windows binary.  Here's related cygwin package versions:
> >     gcc            3.4.4-3
> >     gcc-mingw      20040810-1
> >     gcc-mingw-core 20050522-1
> >     w32api         3.10-1
> >
> > And my build command:
> >     cd vim/src  && \
> >     make ARCH=i686 GUI=yes OLE=yes USEDLL=no \
> >          WINVER=0x0500 -f Make_cyg.mak
> >
> > It seems the same w32api problem has already hit cygwin release.
> > Here's the fix following 7.1.065:
> >
> > *** ../vim-7.1.123/src/Make_cyg.mak   Sun Apr 30 23:28:40 2006
> > --- ./src/Make_cyg.mak        Fri Sep 28 13:25:30 2007
> > ***************
> > *** 502,509 ****
> >   $(OUTDIR)/if_cscope.o:      if_cscope.c $(INCL) if_cscope.h
> >       $(CC) -c $(CFLAGS) if_cscope.c -o $(OUTDIR)/if_cscope.o
> >
> >   $(OUTDIR)/if_ole.o: if_ole.cpp $(INCL)
> > !     $(CC) -c $(CFLAGS) -D__IID_DEFINED__ if_ole.cpp -o $(OUTDIR)/if_ole.o
> >
> >   if_perl.c: if_perl.xs typemap
> >       $(PERL)/bin/perl `cygpath -d $(PERL)/lib/ExtUtils/xsubpp` \
> > --- 502,510 ----
> >   $(OUTDIR)/if_cscope.o:      if_cscope.c $(INCL) if_cscope.h
> >       $(CC) -c $(CFLAGS) if_cscope.c -o $(OUTDIR)/if_cscope.o
> >
> > + # Remove -D__IID_DEFINED__ for newer versions of the w32api
> >   $(OUTDIR)/if_ole.o: if_ole.cpp $(INCL)
> > !     $(CC) -c $(CFLAGS) if_ole.cpp -o $(OUTDIR)/if_ole.o
> >
> >   if_perl.c: if_perl.xs typemap
> >       $(PERL)/bin/perl `cygpath -d $(PERL)/lib/ExtUtils/xsubpp` \
> >
>
> I don't use Cygwin, but I guess for this to work both with old and new
> versions of the API we also need a change in iid_ole.c.  Similar to the
> change for MingW.

Hi, Bram,

I believe the MingW patch has already solved the backward
compatibility problem.  I checked macro definition when I built VIM
with Make_cyg.mak under cygwin, __MINGW32__ is defined, it looks
like MingW is used when build under cygwin.  I don't know the detail
though.  If that's true, backward compatibility should have already
been solved.

In order to be sure, I managed to find an old cygwin installation
with w32api-3.2-1, this version is announced on 2005-1-2, according
to the announcement e-mail here:
    http://sourceware.org/ml/cygwin-announce/2005-01/msg00003.html

It should be old enough :)   I can build VIM successfully with the
above patch.  I also tried w32api-3.7-1, no problem found.

I remember the build problem begins about a month ago (about the
same time when build problem with ActivePerl 5.8.8 happened).  After
I check my cygwin update log, I think the failure should happen
after upgrade to w32api-3.10-1.  w32api release announcement also
confirmed my guess:
    http://sourceware.org/ml/cygwin-announce/2007-08/msg00002.html

One of the change is:
    [mingw-Bugs-1751565]
    * include/basetyps.h (IID); Guard with __IID_DEFINED__.

Hope that helps.

My Best Regards!

-- 
Guopeng

--~--~---------~--~----~------------~-------~--~----~
You received this message from the "vim_dev" maillist.
For more information, visit http://www.vim.org/maillist.php
-~----------~----~----~----~------~----~------~--~---

Raspunde prin e-mail lui