[ansi-test-devel] Running tests out of place

[Juan Jose Garcia-Ripoll]

I forward this again, given that the mailing list maintainers do not care
to approve my previous post. The original attachments can be found here

https://docs.google.com/open?id=0B2brKGbNKZYDNTU0YjM4ZTYtMDk4OS00OWNlLTkzN2YtMGM0YmE0OGZmY2Vi

On Thu, Dec 8, 2011 at 9:07 PM, Juan Jose Garcia-Ripoll <
juanjose.garciaripoll at googlemail.com> wrote:

> Hi, I have been bothered several times by the inability of the ANSI test
> suite to run on a directory different from the one it resides in. This is a
> problem for various reasons:
>
> * The source tree might be read-only. The test suite then is not allowed
> to create compiled files or temporary files.
>
> * Some tests fail and leave temporary directories scattered around. This
> happens quite often. It is hard to clean the mess without keeping a list
> that differentiates the sources from the temporaries.
>
> * If I want to keep state clean, I have to make a copy of the whole
> directory every time I want to run the tests. This happens quite often when
> I have to test multiple copies of ECL with a variety of settings.
>
> Knowing this I have worked on the test suite, changing the tests so that
> they can be loaded from a different directory (what I call the "scratch"
> directory). I attach the changes which consist on this:
>
> * The test suite defines some variables (see doit.lsp) which store the
> location of the sources and the location of the scratch directory in which
> the test suite is going to work.
> * CLTEST: points to the scratch directory.
> * CLTESTSRC: points to the sources.
> * Various tests require some read-only file to be present. Formerly it
> used to be their own test file. I changed this to "ansi-aux.lsp" and
> ensured that such a file is present in the scratch directory.
> * Some tests actually require entire files to be copied (see
> compile-file.lsp and load.lsp)
>
> In addition, I attach an utility script which I use when I want to run
> just one set of tests. It has been customized to work also out of place.
>
> Could you guys please have a look, run it with your favorite
> implementation (I tried SBCL and ECL) and hopefully incorporate the changes?
>
> Juanjo
>
> --
> Instituto de Física Fundamental, CSIC
> c/ Serrano, 113b, Madrid 28006 (Spain)
> http://juanjose.garciaripoll.googlepages.com
>



-- 
Instituto de Física Fundamental, CSIC
c/ Serrano, 113b, Madrid 28006 (Spain)
http://juanjose.garciaripoll.googlepages.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.common-lisp.net/pipermail/ansi-test-devel/attachments/20111210/ba2662f9/attachment.html>