[Openmcl-devel] OpenMCL future
gb at clozure.com
Sat Sep 7 16:55:16 UTC 2002
On Sat, 7 Sep 2002, Sven Van Caekenberghe wrote:
> This test suite (or a version of it) is in CLOCC now, under
> tools/ansi-test. I don't know the current relationship between the
> CLISP and CLOCC source trees, i.e. what happens when new tests or
> platforms (like openmcl) are added, where do you check them in ?
> I took a look at the test suite from CLOCC and ran it in openmcl 0.13.
> The problem is than (too) many tests need their expected results put it
> for each platform. All these failed along the lines of (+ 1 2) returned
> 3 and UNKNOWN was expected ;-). If we want to make this work for
> openmcl, that'll require some work and we need to know where to commit
> these changes to (cfr above).
It's good to have these kinds of things; from time to time, I even
remember to run a slightly modified version of MCL's test suite in
OpenMCL. (If I get any errors - other than ED being undefined, a couple
of math library functions showing slight differences because of the
lack of single-precision versions of these functions under Darwin,
and a couple of things where neither implementation handles case
conversion correctly when translating logical pathnames - I know
that I've broken something.) Sometimes, I even remember to run
through the test suite -before- putting a release together ...
It's also good to have as many of these as possible: CL is a large
language, and everyone programs in a different subset of it. It's
not surprising that every test suite excercises a different subset
of the language.
When MCL 2.0 was released (about 10 years ago), it claimed to be
CLtL2-compliant. I think that that claim was justified, and its
test suite tried to ensure that it was.
I don't have (and don't know where to find) tests that exercise
all of the post-CLtL2 ANSI changes. One of Paul Graham's books
had an appendix which tried to list all of those changes; while
helpful, I don't think that that list was exhaustive.
Putting together such a list (and a test suite derived from it)
would be a fair amount of work. There are some useful starting
points (e.g., check the HyperSpec's list of X3J13 cleanup issues
which passed after CLtL2's publication date).
If I recall correctly (and I can't find email or other evidence
to confirm this, so I apologize if I'm not remembering correctly),
Kent Pitman asked X3J13 for and was granted a certain amount of
editorial license: he was allowed to use his judgement in lots
of cases, rather than require X3J13 to haggle over each obscure
detail. That was certainly expedient, but (if I'm remembering
this correctly) there are things in ANSI CL that no cleanup
issue refers to.
Openmcl-devel mailing list
Openmcl-devel at clozure.com
More information about the Openmcl-devel