[Ecls-list] Fwd: [ANNOUNCE] usocket 0.2.2

[Erik Huelsmann]

On 1/5/07, Juan Jose Garcia-Ripoll <jjgarcia at users.sourceforge.net> wrote:
>
> > I also want to take the opportunity to report a problem I have with
> > ECL: when running the RT tests which come with usocket, a number of
> > tests return a list with all values returned by the forms in the progn
> > of the test, while other lisp implementations only return the last
> > value in the progn. [...]
> > The problem can be found by checking out the usocket code from its
> > Subversion repository. The only tests failing are the ones which
> > exhibit this problem. Looking at the content of the list, the tests
> > would normally (without this bug) have succeeded though.
> > The RT version I used is the one I got with SBCL, I think. Anyway it
> > is unpatched and *not* the one in the ecl repository.
>
> First of all let me tell you that, within my limited time and
> possibilities, I am open to helping people port any kind of libraries
> to ECL and I am happy to fix any problem.

Thanks. I want to do all the work myself when possible, but I don't
think *if* the problem is on the ECL side, I won't be able to.

> Now, as for usocket, I checked it out and, apart from the fact that
> you were not using the rt from SBCL but the one from b9, your library
> does not work for me. Tests do not even run because of this in
> test-usocket.lisp
>
> (defparameter *soc1* (usocket::make-socket :socket :my-socket
>                                            :stream :my-stream))
>
> where make-socket is defined in usocket.lisp as a function taking only
> one keword argument, :socket

Ok. Did you try 0.2.2 or trunk? I ran tests with trunk and - possibly
falsely - assumed you would try trunk too. From what you tell me, I
start to think you may have tried the 0.2.x branch, 0.2.2 tag or the
asdf installable tar.gz.

> (defun make-socket (&key socket)
>   "Create a usocket socket type from implementation specific socket."
>   (make-stream-socket :socket socket))
>
> What you say about rt, though, is strange. This library is also used
> by Paul Dietz's ANSI test suite and it runs perfectly in ECL (I
> routinely use it for testing ECL itself).

I have one version installed with which I test 7 of the 9 supported
lisps (the other 2 are tested by their contributors). I don't have
problems with any of the other lisps. I'm not blaming you, but am
hoping to find out what I'm doing wrong or where the problem is.

Thanks for taking the time to run the tests. If you tested with 0.2.2,
could I ask you to try once more with trunk? Thanks so much in
advance!

bye,

Erik.