Thread: mingw compile error
Hi,
I noticed that if I build postgresql-8.3.1 on mingw with
--disable-shared --without-zlib, I get a compiler error when it goes to
build the regress test in src/test/regress.
The problem is here,
$(NAME)$(DLSUFFIX): $(shlib)
rm -f $(NAME)$(DLSUFFIX)
$(LN_S) $(shlib) $(NAME)$(DLSUFFIX)
It uses the variable $(shlib) which is only defined if the build is
configured with a shared library. Therefor, the ln -s command comes out
like 'ln -s regress.dll', and the build fails.
Bob Rossi
Bob Rossi <bob_rossi@cox.net> writes:
> I noticed that if I build postgresql-8.3.1 on mingw with
> --disable-shared --without-zlib, I get a compiler error when it goes to
> build the regress test in src/test/regress.
That's the least of your problems --- the regression tests wouldn't work
anyway, since they depend on four or five different shared libraries
such as plpgsql.
I'm not sure why we still pretend to support --disable-shared at all...
regards, tom lane
On Thu, May 15, 2008 at 12:37:42PM -0400, Tom Lane wrote: > Bob Rossi <bob_rossi@cox.net> writes: > > I noticed that if I build postgresql-8.3.1 on mingw with > > --disable-shared --without-zlib, I get a compiler error when it goes to > > build the regress test in src/test/regress. > > That's the least of your problems --- the regression tests wouldn't work > anyway, since they depend on four or five different shared libraries > such as plpgsql. > > I'm not sure why we still pretend to support --disable-shared at all... Umm, I don't know what you are talking about. After disabling the building of that directory, everything appeard to work just fine. Do you think I'll have runtime errors because of the --disable-shared option? Bob Rossi
On Thu, May 15, 2008 at 11:18:18AM -0400, Bob Rossi wrote: > Hi, > > I noticed that if I build postgresql-8.3.1 on mingw with > --disable-shared --without-zlib, I get a compiler error when it goes to > build the regress test in src/test/regress. > > The problem is here, > $(NAME)$(DLSUFFIX): $(shlib) > rm -f $(NAME)$(DLSUFFIX) > $(LN_S) $(shlib) $(NAME)$(DLSUFFIX) > It uses the variable $(shlib) which is only defined if the build is > configured with a shared library. Therefor, the ln -s command comes out > like 'ln -s regress.dll', and the build fails. Ping. Anyone care? Bob Rossi
Bob Rossi <bob_rossi@cox.net> writes:
>> I noticed that if I build postgresql-8.3.1 on mingw with
>> --disable-shared --without-zlib, I get a compiler error when it goes to
>> build the regress test in src/test/regress.
> Ping. Anyone care?
Well, no, not a lot --- you haven't explained why it's necessary for
you to build with --disable-shared. Mingw certainly supports shared
libraries, and it's hard to deny that that's the wave of the future,
and as I already mentioned to you there are large areas of Postgres
functionality that never have and never will work without shared
library support.
For the record, so far as I can tell the src/test/regress makefile has
never supported --disable-shared build. If it worked for you before,
it was because the upper-level makefiles didn't propagate the "make all"
request down to src/test/regress. But that started to happen in 8.2,
and yours is the first complaint we've gotten, so it doesn't seem to be
of wide interest.
FWIW, I notice that CVS HEAD doesn't try to build the shlib in
src/test/regress, as a result of unrelated cleanups by Peter E.
It does still try to build the two contrib/spi .so files that the
regression tests want ... I have no idea if that would fail on
your tool chain or not.
regards, tom lane