Subject: Re: Changes to psiconv for Windows
From: Sam TH (sam@uchicago.edu)
Date: Sun Feb 11 2001 - 12:18:25 CST
On Sun, Feb 11, 2001 at 01:14:26AM -0500, Tom Briggs wrote:
>
> > > I've finally gotten psiconv building correctly on Windows. I had to make
> > > a few simple changes to compat/Makefile.in, psiconv/Makefile.in and ltconfig
> > > in order to get it to work; unfortunately I don't have a diff handy, but the
> > > changes are extremely simple and should be easy enough to locate in Abi's
> > > Bonsai.
> >
> > I hope you are saying: compat/Makefile.am and psiconv/Makefile.am. If
> > not, these changes will get lost next time I run automake :-(
> > I sent a patch a few minutes ago; can you please check whether it
> > contains valid definitions for you? If not, can you please send
> > me your changes, so that I can incorporate them somehow into the
> > Makefile.am?
>
> The problem is that Makefile.in (and therefore Makefile) contains gcc
> specific preprocessor flags (-Wp,...). This causes the resulting
> Makefile to be unusable with MSVC.
> I looked at both Makefile.am as well as the patch you mentioned, and neither
> seemed to contain the troublesome flags. Therefore I doubt that the patch will
> fix anything.
The nasty dependency generation stuff (which is what is causing the
problems here) is generated by automake automatically. The way around
it is to pass the -i argument to automake when you run it. Then the
stuff isn't generated.
Do you think you can do that, Frodo?
>
> > > 2. Is there any way to tell configure to not compile in the compat stuff?
> > > As far as I can tell none of that stuff is necessary for Windows, but having
> > > it present makes linking a little uglier. This wasn't hard to work around
> > > either, but it would be better for the Win32 build if the compat stuff
> > > wasn't compiled in at all.
> >
> > It is really needed for some systems. It would be quite hard to not
> > add it. If libtool handles msvc correctly, it should work automatically;
>
> I understand that it's needed by some systems, but I see no reason why it
> has to be included for those that don't. It caused some easily fixable problems
> for me with MSVC, but on a fundamental level, I object to not only compiling in
> but hacking Makefiles to handle code that I know isn't needed.
> I'm not saying that the need (or lack of a need) for the compat stuff
> should be automagically detected - a simple --without-compat option or
> something like that would be satisfactory as far as I'm concerned.
Writing a little option like this should be *really* easy, so that
compat isn't included in subdirs in the Makefile.am at the top level.
If you don't want to do it Frodo, I can.
sam th
sam@uchicago.edu
http://www.abisource.com/~sam/
GnuPG Key:
http://www.abisource.com/~sam/key
This archive was generated by hypermail 2b25 : Sun Feb 11 2001 - 12:16:52 CST