clisp packaging problem

Reini Urban
Tue May 3 14:30:00 GMT 2011

[followup on cygwin]

2011/5/3 Ken Brown:
> The clisp package has a problem very similar to the libsigsegv problem
> reported by Corinna
> (  I'm starting a
> new thread because it's a different package, but the basic issue is the
> same: The build is set up to treat Cygwin as being "like Windows" rather
> than being a POSIX system.
> I recently had occasion to build clisp-2.49 and ran into trouble because the
> build on Cygwin links against various libraries in /usr/lib/w32api.  See
> for details.  The problem can be traced to the following in
> clisp-2.49/modules/syscalls/
> if test "$ac_cv_header_shlobj_h" = yes ; then
>   LIBS=${LIBS}' -luser32 -lole32 -loleaut32 -luuid';
> fi

Unfortunately we need those libs for some minor calls into the winapi.
But there's no other conflict then our official libuuid, which is prefered
over the win32api uuid.a without my patch.

I never uploaded 2.49 because the are more severe issues with dynamic modules
besides those minor patches.

> together with the following in clisp-2.49/modules/syscalls/calls.c:
> #if defined(WIN32_NATIVE) || defined(UNIX_CYGWIN32)
> #include <shlobj.h>

Unfortunately we need this also.

> These are only two examples.  Here's a third I noticed: The following lines
> in clisp-2.49/src/encoding.d make the Cygwin build of clisp use DOS line
> endings:
>  #if defined(WIN32) || (defined(UNIX) && (O_BINARY != 0))
>  pushSTACK(S(Kdos));           /* :line-terminator */

Of course I patched that. With CRLF we cannot pipe into clisp.

> This one is minor, but I think it indicates a basic misunderstanding of what
> Cygwin is.

I believe the latter.

Problem reports:
Unsubscribe info:

More information about the Cygwin mailing list