[asdf-devel] Latest version has multiple test failures

[Faré]

On Mon, Oct 14, 2013 at 4:55 PM, Robert P. Goldman <rpgoldman at sift.info> wrote:
> * ECL fails test-stamp-propagation:
>
> TEST ABORTED: These two expressions fail comparison with EQUAL:
>  (FILE-WRITE-DATE FILE) evaluates to 3590769832
>  STAMP evaluates to 3590766232
>
Weird. That's a 1h difference, and sounds like a problem that ECL
would have with daylight saving time or such on Mac. It works on
Linux.

I modified the script to not be sensitive anymore to discrepancies
between Lisp time and system time. This fixes the test script on
SBCL/Windows, and hopefully also for ECL/Mac.

Note that this is a new test I've just added, that demonstrates THE
defsystem bug that has plagued about every free software CL defsystem
since forever until ASDF 3. I thought the Franz defsystem was safe,
but apparently, it still has a smaller version of the bug; LispWorks,
on the contrary, seems to pass the test, where my memory told me it
didn't; maybe I was confused between the two. I haven't tested Genera
for the smaller bug.

> * Allegro "modern" mlisp fails fourteen tests:
>
Oh, dang, that's a test infrastructure bug: in the hopes of avoiding
issues on Windows, I modified the output-translations to use the short
:implementation-type name instead of :implementation, and this causes
FASL clashes between mlisp and alisp. I just restored the previous
behavior. It looks like this wasn't the cause of my Windows woes
either, after all.

> * lispworks fails test-stamp-propagation on both linux and Mac OS X.
>
Yes, it failed to fail. I marked the LispWorks defsystem as passing
the test rather than failing. (The script tests the native defsystem
as well as ASDF, on Allegro, LispWorks, Genera.)

> Unfortunately, I am under multiple deadlines, and will have little time
> to investigate these in the next week.
>
Sorry for the mess. The last few days, Dave Cooper offered to help
testing on Windows,
and I've made lots of small changes to the test system and fixed a few
UIOP and ASDF bugs,
not all of them exclusive to Windows.

> But clearly we cannot release 3.0.3 under these circumstances.
>
Indeed.

—♯ƒ • François-René ÐVB Rideau •Reflection&Cybernethics• http://fare.tunes.org
Don't have good ideas if you aren't willing to be responsible for them.
	— Alan Perlis