[Bordeaux-threads-devel] test failures

[Anton Vodonosov]

28.02.2012, 04:32, "Vladimir Sedach" <vsedach at gmail.com>:
> First I want to say that this is the first I've heard of the
> cl-test-grid project, and it is really cool. You should blog about it
> and let more people know.

Hello. Thanks for the positive judgment. The project is in quite early stage, 
there are things to do to make it more useful. But I already reported about this 
project I think to all the mailing lists of the libraries having test failures.

> About testing CLISP and bordeaux-threads, is there a way to test and
> report results for a custom compiled version of CLISP? It's had thread
> support for some time now, but needs to be compiled with it turned on.

If you want to run the tests yourself, it's easy, the instructions are on the
github page. As of today, CLISP is unable to submit test results
to the server automatically, because of broken drakma; the easiest way
to work this around is to use another lisp (e.g. SBCL, CCL) after the
CLISP tests are completed, to submit the results. An example how
to do this is at the last line of run.sh.sample.

If you can't afford yourself to run the tests, then I will keep in mind the possibility
to run bordeaux-threads on multi-threaded CLISP. I am subscribed to the
clisp-devel and was aware that the multi-threading is almost completed,
except for synchronizing internal CLISP data structures; but somehow
missed the point when it shaped to clearly specified constraint: perform
all defclass, defmethod and other CLOS definitions before spawning
threads. Also I wasn't aware that bordeaux-threads already supports
CLISP. So, thanks for the information.

I can't switch to updating my CLISP immediately, because there are
relatively simple, but very worthwhile things to implement in 
cl-test-grid: I am finishing a function to automatically 
find regressions between two quicklisp releases (would be
also good to search for regressions between to versions of
the same compiler, but I'll leave it for latter); next it's necessary
to collect result not as just ok/fail for a whole library, but detailed to 
particular failed tests - otherwise we can't distinguish regression
between library version where 1 test fails and 3 tests fails; this
is already done for the tests based on RT framework; 7 more
frameworks to do; then it's necessary to add more libraries;
then I want to make running the tests more convenient... 
So, as I said, I will keep in mind the possibility to use multithreaded CLISP, 
but will postpone it to the next time when I will update CLISP on my machine.


Is there anyone who have time to check the bordeaux-threads
failed tests? Is it a bug in tests?

Best regards,
- Anton