Mailing-List: contact cygwin-help AT cygwin DOT com; run by ezmlm
Sender: cygwin-owner AT cygwin DOT com
Mail-Followup-To: cygwin AT cygwin DOT com
Date: Sun, 2 Jun 2002 00:47:47 -0400
From: rich-paul AT rich-paul DOT net
To: cygwin AT cygwin DOT com
Subject: Re: Building cygwin1.dll
Message-ID: <20020602004747.A10315@monster.rich-paul.net>
References: <000001c204b3$f2b01bc0$0200a8c0 AT sknet01> <000001c204b3$f2b01bc0$0200a8c0 AT sknet01> <20020528120346 DOT B21064 AT monster DOT rich-paul DOT net> <4 DOT 3 DOT 1 DOT 2 DOT 20020528122339 DOT 0260f7f0 AT pop DOT ma DOT ultranet DOT com> <20020528190238 DOT A22088 AT monster DOT rich-paul DOT net> <20020528232930 DOT GA18281 AT redhat DOT com> <20020529091728 DOT D22088 AT monster DOT rich-paul DOT net> <20020529143618 DOT GA2808 AT redhat DOT com>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
User-Agent: Mutt/1.3.16i
In-Reply-To: <20020529143618.GA2808@redhat.com>; from cgf-cygwin@cygwin.com on Wed, May 29, 2002 at 10:36:18AM -0400

OK, why don't I start over:

	My assumption, from the previous discussion, was that the directions
	in the FAQ were, in theory, sufficient to actually perform the task.
	Therefore, I assumed that cygwin-1.3.10-1-src.tar.bz2 was supposed
	to be able to build alone, and that it's inability to do so was
	brokenness on it's part, which has been hidden because it's
	generally only compiled as part of the larger 'net-release',
	which contains other things.

	This was bolstered by being able to compile a working cygwin1.dll
	using, I assume, my installed version of w32api in place of the one
	that belongs in the source tree.  It may be, however, that doing
	this worked "by accident" and should not be assumed to be a workable
	thing in the future.

	But if there are other packages that *SHOULD* be required to build
	cygwin1.dll, then my assumption was invalid.  My next instict was to
	say "OK, the brokenness is in the FAQ".  But having re-read that
	entry, I cannot say that it's technically wrong.  It merely says to
	put "the sources" into /src.  It does not say what constitutes "The
	sources".  It was *my* assumption that the only sources required to
	build cygwin1.dll were contained in cygwin-1.3.10-1-src.tar.bz2.

	So now the question becomes, what is the minimum subset of the
	source distribution that must be unpacked into /src in order for
	following the directions in the FAQ to work, and how should they be
	put there.  I assume that the three empty directories, w32api, mingw,
	and cinstall should be filled, so I'll start with them and see if
	that produces a dll.

	Another question:  Are there any directions as to how to use the
	mknetrel package from the cygwin-apps cvs tree?  I've read some of
	the script, and it looks like it might make life easier, but it
	appears to need things mounted in particular absolute directories.
	If there are no such docs, would you be interested in having some?

On Wed, May 29, 2002 at 10:36:18AM -0400, Christopher Faylor wrote:
> On Wed, May 29, 2002 at 09:17:28AM -0400, rich-paul AT rich-paul DOT net wrote:
> >	I noticed that removing that directory exposes another problem,
> >	the makefile wants to make a target called w32api, but there are
> >	no rules for such a target.  Of course, make hapilly thought
> >	it had already made it, in the form of a directory, so didn't
> >	complain.
> 
> That's because you *need* the directory to be populated.
> 
> >	I removed the dependencies from four lines, and am trying again
> >	...
> 
> Wrong "solution".
> 
> >	Hmmm ... switching the lib directories to use the installed
> >	mingw32/win32api
> 
> Wrong "solution".
> 
> >	OK, got farther, added a check for the cross-directory
> >	libkernel32.a dependency ....
> >
> >	At this point, started dying of 'parse error before symbol
> >	__gnuc_va_list' in stdio.h.  It does include stdarg.h, but
> >	perhaps whatever keeps stdio.h from exposing va_list, as the
> >	comment in the gnuc stdio says, is the culpret.  Anyway, I didn't
> >	want to dig too deeply, so I played with making .i files  for
> >	a while, and resorted to commenting out the conditionals at
> >	line 143, so that __VALIST would be #defined to char*.
> 
> Something is wrong in your installation.
> 
> When, you think about this problem, think about it from the point of
> view that many people (me, for instance) are building cygwin.  We don't
> build it by hacking at the Makefile.  We build it by having the complete
> sources installed.
> 
> If you have to start editing things, you're obviously going down the
> wrong track.
> 
> cgf
> 
> --
> Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
> Bug reporting:         http://cygwin.com/bugs.html
> Documentation:         http://cygwin.com/docs.html
> FAQ:                   http://cygwin.com/faq/
> 

-- 
Got freedom?  Vote Libertarian:  http://www.lp.org

--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Bug reporting:         http://cygwin.com/bugs.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/