Re: [pure-lang-users] linux PPC

[Albert G. <Dr....@t-...>]

Tim Haynes wrote:
> I'm also attaching a trivial patch to Makefile that should remove logs
> before `make check', thereby starting afresh at test 1 every time the way I
> like it.

No, that's not the way those tests are supposed to work. I know that
this confusion arises because the test logs aren't included in svn right
now. Maybe I should do that, but I made it a strong habit to never put
generated stuff in a repository.

But normally the output of those regression tests is *only* to be
regenerated when the tests themselves change (this is done
automatically), otherwise the whole idea of the check is to *compare*
your output against the *existing* logs, which presumably were generated
for you on a system where it's already known that the interpreter works
correctly.

The only circumstances under which I regenerate the logs from scratch is
when there's some change in the compiler affecting the output of -v7. In
such a case I just run 'make cleanlogs check', *after* carefully
checking the diffs against the existing logs and making sure that
everything still works.

Also note that the output of -v7 is truly platform-independent, it's
only abstract code (essentially rewriting rules annotated with de Bruijn
indices and descriptions of the corresponding pattern matching automata)
which isn't supposed to change. It doesn't even include the generated
LLVM assembler code, because that code is subject to change (e.g., if
there's a bug in the code generator, or some new semantics-preserving
optimizations are applied). Whereas the output of -v7 is just the
abstract "meaning" of your program.

Ok, I hope that this clears things up a bit. :) I'll probably have to
put some notes about this into the README file.

Albert

-- 
Dr. Albert Gr"af
Dept. of Music-Informatics, University of Mainz, Germany
Email:  Dr....@t-..., ag...@mu...
WWW:    http://www.musikinformatik.uni-mainz.de/ag