RE: [INTERFACES] The problem between libpq and libpgtcl for WIN32

Hi all.

Hiroki Kataoka wrote
> Hello all.
>
> > -----Original Message-----
> > From: owner-pgsql-interfaces(at)postgreSQL(dot)org
> > [mailto:owner-pgsql-interfaces(at)postgreSQL(dot)org]On Behalf Of Hiroshi Inoue
> > Subject: [INTERFACES] The problem between libpq and libpgtcl for WIN32
> >
> >
> > The treatment of pgresStatus[] in libpq must be changed ?
> >
> > In case of Windows,global data interface between dll's is
> > different from UNIX.
> > If fact,libpgtcl references global data pgresStatus[] in libpq
> > and pg_result -status command does not work properly for
> > Windows.
> >
> > I think it is preferable to use function interfaces than gobal
> > data references.
> >
> > Any ideas ?
>
> I solved this problem by the change in the following.
>
> ======== cut here ========
> *** libpq-fe.h.dist Tue Oct 6 23:16:48 1998
> - --- libpq-fe.h Tue Dec 15 03:52:34 1998
> ***************
> *** 50,56 ****
> - --- 50,60 ----
> } ExecStatusType;
>
> /* String descriptions of the ExecStatusTypes */
> + #ifdef WIN32
> + __declspec(dllimport) const char * const pgresStatus[];
> + #else
> extern const char * const pgresStatus[];
> + #endif
>
> /* PGconn encapsulates a connection to the backend.
> * The contents of this struct are not supposed to be known to
> applications.
> *** libpqdll.def.dist Tue Oct 6 23:05:02 1998
> - --- libpqdll.def Tue Dec 15 03:52:44 1998
> ***************
> *** 63,68 ****
> lo_unlink @ 60
> lo_import @ 61
> lo_export @ 62
> ! pgresStatus @ 63
> PQmblen @ 64
> PQresultErrorMessage @ 65
> - --- 63,68 ----
> lo_unlink @ 60
> lo_import @ 61
> lo_export @ 62
> ! pgresStatus @ 63 DATA
> PQmblen @ 64
> PQresultErrorMessage @ 65
> ======== cut here ========
>
> "pg_result -status" operates correctly by this method. Moreover, it is
> thought that this method is easy because the change part is in libpq.
>

The word _declspec(dllimport) is Microsoft-specific and other compilers
can't handle this keyword ?

Magnus, libpq for WIN32 is only for Microsoft Visual C++ ?
(Seems makefiles win32.mak are for Microsoft Visual C++).

For this problem,Tom Lane wrote

> But it seems reasonable to add a corresponding function call, say
>
> const char * PQresStatus (ExecStatusType status);
>
> and change libpgtcl to use that instead. We could even deprecate

I agree with him because I don't like global data references.

But the way Hiroki proposed has a advantage that binaries(libpq,libpgtcl)
are invariant except the dll's for WIN32.

Thanks.

Hiroshi Inoue
Inoue(at)tpf(dot)co(dot)jp