Re: [Jython-dev] bugtests

[Jeff A. <ja...@fa...>]

Thanks for the quick reply Jim. I'll feel no compunction then about 
removing further tests if they're more trouble than they're worth.

And perhaps the rest (and the ant target), although harmless, should be 
consigned to history. I'm just a bit wary since the release instructions 
called for it, although this too may be old text.

I spotted that driver.py and support.py provide for calling jythonc, but 
I haven't seen it attempted in the surviving test. I didn't know why we 
dropped jythonc. I've mused (without looking) on how one implements 
generators, realising you somehow have to achieve a resumable function.

Jeff Allen

On 28/08/2019 22:30, Jim Baker wrote:
> I would remove them completely. These tests are largely 18 or so years 
> old, and date to the initial implementation of Jython/JPython. I 
> personally surprised they are still there, but there is always cleanup 
> to be done in such a large project, and we just didn't around to this 
> specific cleaning in the past.
>
> There is also some testing of JPythonC, which we abandoned 10+ years 
> ago, since it is not possible to implement such key functionality as 
> generators, etc, strictly in Java instead of Java bytecode - at least 
> not without significant performance hits (eg using a Python bytecode 
> approach).
>
> - Jim
>
> On Wed, Aug 28, 2019 at 3:19 PM Jeff Allen <ja...@fa... 
> <mailto:ja...@fa...>> wrote:
>
>     I have not paid any attention to the bugtest ant target but Frank's
>     notes on releasing say one should run it. It fails miserably.
>
>     The tests fail on Windows, because of the way the command line is
>     created in bugtests/support.py, but if I fix that, most pass, but
>     many
>     do not, and for what appears at first to be a good reason.
>
>     Maybe some of these failures are regressions, but on the whole, the
>     problem seems to be with the tests. test235 and test236 need .java
>     files
>     missing from bugtests/classes; test321 and quite some others fail
>     because the class path does not include dist/javalib/*; yet others
>     are
>     more obscure.
>
>     There has been a great culls in the past. This changeset
>     (https://hg.python.org/jython/rev/f494dce0a7e5) is an example and it
>     looks carefully reasoned -- tests are superseded by regrtests.
>     However,
>     it accidently blows away some files used by tests it leaves
>     behind. The
>     bugtest target can't have run cleanly since that point, which is
>     an odd
>     way to leave things.
>
>     Do we take these tests seriously?
>
>     -- 
>     Jeff Allen
>
>
>
>     _______________________________________________
>     Jython-dev mailing list
>     Jyt...@li...
>     <mailto:Jyt...@li...>
>     https://lists.sourceforge.net/lists/listinfo/jython-dev
>