[Texmacs-dev] nogencc-11 and TeXmacs-misc

[David Allouche]

New this version:

    Synched with TeXmacs-1.0.0.5.

    Restored H convenience macro in hashmap.cc and
    hashmap_extra.cc because Joris has feelings against that
    change.

    Added a bit of makefile logic to avoid using the dirty (and
    stupid) "src/makefile-nodeps" trick. Dependences analysis is
    no longer required for compilation.

    Improved factorization of src/makefile.

Tarballs are available at;
http://ddaa10.free.fr/texmacs/patch/1.0.0.5


I had a look at the TeXmacs-misc stuff. It is very problematic, because 
the use of the -frepo option is not really compatible with the library 
based build used previously, and on which the TeXmacs-misc build depends.

I have selected two solutions which are not too ugly:

 1. Use implicit (and duplicate) template instanciation for static
    libraries. That way the libraries would be readily usable by
    testing tools, but they will not be useful for building TeXmacs,
    and will have to be built in a separate object directory.

 2. Let the TeXmacs-misc makefiles recompile the TeXmacs objects with
    template repositories (-frepo). Since the repository is different
    for each test program, each will have to be compiled in a
    different Object directory. The global compile time would be
    greater, but individual incremental recompilation would generally
    be faster. Individual incremental recompilation is what is
    critical to the developper.

If the solutions 2 is chosen, it would be much simpler to incorporate
the TeXmacs-misc source tree in the main TeXmacs source tree. That way
there would be not need to either:

 -- Duplicate makefile code in TeXmacs and TeXmacs-misc for
    object aggregation and include paths of TeXmacs objects. That
    approach would make the TeXmacs-misc distribution more fragile by
    creating hidden dependences.

 -- Including the TeXmacs/src makefile in the TeXmacs-misc/src
    makefile. Which would require additional complication in the
    former to make it location independant, or hacks in the latter to
    let included rules operate in an alien context.

Is there any good reason not to incorporate TeXmacs-misc in the main 
TeXmacs source distribution, apart for avoiding a 7% increase in archive 
size?

Not only it would make my job simpler, but it would also considerably 
increase the visibility of these test tools. That is important because 
that would probably make it easier for most of us to understand those 
parts of TeXmacs.
-- 

                                  -- David --