[Ecls-list] ECL support for cl-launch and xcvb

Faré fahree at gmail.com
Mon Nov 9 11:16:57 UTC 2009


Dear Juan,

> ECL does not dump images.
Yes, sorry this was a misnomer for linking. cl-launch and xcvb try to
offer a uniform declarative interface to building software across all
CL implementations, so I'm using the same word even though it means
something very different with ECL.

> As such, there is no way that si:getenv
> returns the environment from a previous session. If you get NIL then
> the environment variable is not defined or there is something wrong
> going on with the POSIX function to retrieve the environment variables
>
I'll see if that bug is still there if/when I'm past the other bug.
Interestingly, the si:getenv function worked at the command-line; it
just failed badly when compiled into compute-arguments.

>> With 9.10.2, I fail at test 73 when I first try to dump an image and
>> resume from it. Apparently, the primitive (si:quit 0) fails miserably,
>
> Yes, this was fixed partly yesterday night and partly today.
>
Thanks a lot!

>> cl-launch has a love-hate relationship with ECL. ECL was the first
>> implementation for which I could produce reliable standalone
>> executables. But it is also an implementation where weird things
>> happen, and features are randomly broken from one revision to the
>> next.
>
> There is probabaly a fundamental difference between ECL and other
> implementations. Namely, the low level details are still moving. From
> the release you were most happy with till today we had to add support
> for five new platforms and chipsets, and this involves changing also
> the way things work at the C level.
>
> So it is not that I personally struggle to break things at random
> places, but the fact is that random things have to be changed because
> now this ARM compiler does not like floating point exceptions, that
> Sun compiler has only partial support for that feature, etc. Your
> library, cl-launch, is also dealing with pretty sensitive stuff --
> executable building -- and thus it is more likely that you hit those
> rough spots.
>
Yes, I understand all that, and I think you're doing a great job. I am
a bit frustrated perhaps that I am handling bits that seem to not have
tests in your regression suite. Hopefully, once made stable, my
programs themselves could be included in such suite, and/or I could
automatically update and test ecl with my programs, and things will go
smoothly from there.

>> I'd like to support it in cl-launch and xcvb, but I am currently at
>> loss at what approach to use to debug such issues.
>
> Faré, you say that you found a problem with 9.8. That is three months
> ago, but I have not read ANY email from you reporting that here. If
> you do not report problems, how am I supposed to find them out? Pester
> us with ANY SINGLE PROBLEM you find with cl-launch and me or somebody
> else contributing here may help in solving it.
>
Up until now, I was tracking ECL through debian, which moves very
slowly. ECL moves very (impressively!) fast, and I will try to follow
it through clbuild instead that seems to be working from CVS and doing
the configuration steps for me (last time I tried to configure and
compile from CVS myself, I failed). I would like to add ECL amongst my
actively supported implementations, and I will.

Thanks a lot for your support!

[ François-René ÐVB Rideau | Reflection&Cybernethics | http://fare.tunes.org ]
Young man, in mathematics you don't understand things,
you just get used to them. -- John von Neumann (1903-1957)




More information about the ecl-devel mailing list