test_struct.py fails when building Python 2.2 on RedHat 6.2/alpha

Tim Peters tim.one at comcast.net
Fri Mar 29 15:17:48 EST 2002


[Mike Carifio]
> I've recently built Python 2.2 on a RedHat 6.2/alpha box.
> When I run the regression test suite ('make test'), test_struct.py
> fails with a floating point exception and stops the regression test
> suite.
>
> Some questions:
>
> 1) Is this a know error or have I discovered something new?

It's news to me.

> I tried to find out if this error was seen before by going to
> the python bugs database ...
> and searching the newsgroup in Google. Are there other sources?

You already found the best.

> I didn't find a good way to search the sourceforge bug tracker for
> specific keywords (like 'test_struct' or 'alpha').

There's a "Search" box in the left margin.  Be sure the dropdown list has
"Bugs" selected.  It turns up 3 test_struct bug reports, all closed.

> 2) Once I take test_struct.py out of the test list, the suite runs to
> completion, but I still have 24 tests that have failed.

Sounds like your installation is a mess.  Open a bug report and attach some
details (like *which* tests failed).

> Are all tests the same?

No.

> For  example, if test_bsddb fails, is my installation bogus or
> should I just shy away from db files on RH6.2/alpha?

It's impossible to guess from what you've said so far -- open a bug report
and people will help there.

> Is there a document or file somewhere that tells me how to interpret
> these results?

No.  The tests are all written in Python, so ambitious <wink> users dig into
them and figure out why they fail.

> 3) And related to (2), is test_struct.py a "mandatory" test?

It tests the struct module.  If you never use the struct module, it probably
doesn't matter whether it fails.  But there's no good reason for it failing,
so it suggests deep problems.  It *sounds* like your build was such a
disaster that you've got severe problems.

> I guess the question I'm asking is: if some of the tests fail, is
> the build to be distrusted? If there's a judgment call to make here, how
> do I inform my judgement?

Sorry, that's up to you.  I personally would not trust a box where any test
failed; tests being skipped is a different issue.





More information about the Python-list mailing list