cygwin/cross patch#2 tetex-2.96.1.20040322

[sorry for resend] I've seen so replies and the archive seems to be up
again.

Date: Sun, 04 Apr 2004 18:49:53 +0200


Olaf Weber writes:

> Taken together, I think the patch is useful starting point when
> modifying the web2c sources for cross-compiling, but I'm not going to
> apply it in its present form.

That's very promising, so I decided to spend a day or so to fix the
major problems and incorporate most of your wishes.  The patch below
is tested and produces fine binary package for Cygwin.

Major attractions are:

 * Cleaner cross compilation makefile usage.  For all binaries that
   are needed during build, the -for-build suffix has been dropped.
   The -for-build suffix for object files (and proglib library) is
   used when cross compiling only.

 * Use of ctangelboot, otangleboot, tangleboot and tieboot during
   build; drop these from cross.ac detection.
   
 * Disentangle fixwrites, splitup, and web2c helpers from proglib and
   kpathsea.

 * Cleanup of DOSISH definition and usage in scripts.

 * Be more careful with *DLL macro definitions.
 
   ***Note for Windows users***: when not using the distributed
   Makefiles, you should now:

     - when building the DLL: #define KPSEDLL_EXPORT __declspec(dllexport)
     - when using the DLL: #define KPSEDLL __declspec(dllimport)

   and similar for TEXDLL*.  Please test, see also kpathsea/cproto.h.

 * Remove `off-patches' (check/trip targets, newly introduce CXXHACK* etc).

 * Some $(srcdir)/VPATH build fixes.

I have tried your KPSE_CROSS_PATH_PROG macro, but I propose to
postpone this until after the autoconf2.50 conversion.

One major point that I have not changed/addressed is:

   6 - Esp. for teTeX formats/bases/mems can be expected to be generated on
       the target host using fmtutil, or simply be copied from a current
       teTeX installation on the build host (source of the libkpathsea,
       required as per 1, above).  So all the makefile targets for
       generating formats/dumps/mems etc should simply be disabled.

because that involves new functionality that I'm not familiar with, and
it breaks the Cygwin packaging so I'd have to revert that anyway.

Because juggling with changesets is not something that CVS (which I've
been using to track upstream) is really helpful with, I'm now sending
cross and Cygwin changes in one patch.  I think it won't be too
disturbing for the discussion.

Greetings,
Jan.



-- 
Jan Nieuwenhuizen <[email protected]> | GNU LilyPond - The music typesetter
http://www.xs4all.nl/~jantien       | http://www.lilypond.org