This is the mail archive of the
mailing list for the Cygwin project.
Re: Please Upload: octave-2.1.71-1
- From: "James R. Phillips" <antiskid56-cygwin at yahoo dot com>
- To: cygwin-apps at cygwin dot com
- Date: Wed, 6 Jul 2005 14:55:46 -0700 (PDT)
- Subject: Re: Please Upload: octave-2.1.71-1
- Reply-to: antiskid56-cygwin at yahoo dot com
--- Igor Pechtchanski wrote:
> > In building the program, the gcc linker commands use the flag
> > "-lreadline". Since both static and dynamic readline link libraries are
> > present in /lib, apparently the linker prefers the dynamic library,
> > ending in .dll.a. Is this a correct understanding of gcc on cygwin?
> Yes, unless the library is explicitly specified or the -non_shared
> (-static) flag is used.
Interesting. Upstream changelog indicates that cygwin specific changes were
made to the configure script some years back, to use -lreadline. My guess is
that the library search during configure found and preferred the static library
libreadline.a without this change.
That is the behavior I observed with the configure script relative to the
lapack library. The way I forced it to not use a static library was to not
provide one, but instead to make liblapack.a a symlink to liblapack.dll.a. By
itself the configure script did not find the lapack library if only
liblapack.dll.a was present.
I wonder if there are autoconf macros to inform configure that files ending in
dll.a are export libraries for dll's, and satisfy library dependencies? If
there are, it might be better for upstream to incorporate these. I know next
to nothing about autoconf.