[Ecls-list] News about release and more
Gabriel Dos Reis
gdr at integrable-solutions.net
Wed Apr 1 18:25:52 UTC 2009
On Wed, Apr 1, 2009 at 11:31 AM, Juan Jose Garcia-Ripoll
<juanjose.garciaripoll at googlemail.com> wrote:
> I have fixed a lot of typos affecting different parts of ECL: handling
> of C streams, compiling certain global function calls, compilation of
> FSET, etc.
> Right now ECL seems to be able to build OpenAxiom, though I really do
> not know what to do with it :-) If you pass me a way to run tests and
> timings of OpenAxiom I will gladly add them to the database of
> automated builds that my computer performs and you will be able to
> monitor the status of both's programs compatibility.
Many thanks for the kind offer. After the build, if you do
you will get the (currently minimal) testsuite run.
OpenAxiom has a test system that currently involves too much human
intelligence, but I'm looking for alternatives -- my other experience
are with DejaGNu and QMTest. I'm not satisfied with either.
What kind of timing are you looking for?
> There has been a major reorganization in the code for building
> optional modules, and in particular ECL can now be shipped in a mode
> that does not require a C compiler.
> First of all, modules now can be built as loadable files, as built in
> extensions that form part of the shared library or not built at all.
> For instance, if you want to add ASDF support you may choose between
> --with-asdf=yes (default), --with-asdf=builtin (no need to load the
> module) and --without-asdf or --with-asdf=no (do not build it).
> A new module that we now have is BYTECMP. This module uses the
> bytecodes interpreter to produce text files that have been
> precompiled. This clever trick of using ECL's minimal compilation
> system allows you to ship ECL as a standalone program where COMPILE
> and COMPILE-FILE work, but which will not have the full speed of a
> native compiler.
> The most interesting use of this is combining flags in the following form
> --disable-shared --with-bytecmp=builtin --with-cmp=no
> The resulting statically linked file will be a full lisp image which
> supports compilation to bytecode and, as I said, does not require a C
> compiler around. Unfortunately bytecodes compilation means C-INLINE
> and some other FFI forms no longer work. Furthermore, the bytecodes
> compiler currently does not support exporting compiled code with
> objects that do not have a printed representation. This should be
> worked out in a near future (help welcome).
> If everybody agrees, I will prepare a release in a couple of days.
The new ECL-based OpenAxiom build is progressing far beyond
the last reported failure point, so I think you have green light as
as far as OpenAxiom is concerned.
Many thanks for your effort.
More information about the ecl-devel