Re: [Jython-dev] 2.2rc1 later this week

[Alan K. <jyt...@xh...>]

[Oti]
 >     [java] test_poll
 >     [java] test test_poll crashed -- exceptions.AttributeError: class
 > 'org.python.modules.os' has no attribute 'pipe'

As I mentioned before, there is no point in running these test_poll: it 
tests for cpython facilities that are not currently available on jython. 
Test_poll should either

A: Be excluded from the test run, or
B: Left in place to remind us that although pipes are not currently 
supported on jython, they theoretically could be, as long as there is a 
subprocess running on the other end of the pipe. This would require 
substantial modification of the jython popen* calls, to use 
Pipe.SinkChannel and Pipe.SourceChannel.

[Oti]
 >     [java] test_select
 > <snipped ok messages>
 >     [java] test test_select produced unexpected output:
 >     [java]
 > **********************************************************************
 >     [java] *** line 2 of actual output doesn't appear in expected
 > output after line 1:
 >     [java] + Unable to run testSocketRegisteredBeforeConnected: no
 > server on port 80
 >     [java]
 > **********************************************************************

This is a simple problem, but I'm not sure how best to fix this test.

The test in question expects to connect to some TCP server on localhost. 
As I always have a HTTP server running on port 80 on my machine, and 
expect most other machines to be the same, I used port 80.

But there is obviously no server running on port 80 on the test machine.

There are three solutions

1. Use a TCP port that *will* always be listening when the test is run; 
the protocol doesn't matter, it just needs to be accepting connections.

2. Run up a TCP server, as a part of the test, solely for the duration 
of the test, then tear it down again. A lot of work for a small gain.

3. Pick a well known HTTP server on the 'net, e.g. www.python.org:80. 
But this could give rise to network latencies that might cause the test 
to fail unnecessarily.

What think ye? Are there any existing TCP servers running on the test 
machine?

[Oti]
 >     [java] test_socket
 > <snipped ok messages>
 >     [java]
 > ----------------------------------------------------------------------
 >     [java] Ran 67 tests in 2.609s
 >     [java] OK
 >     [java] test test_socket produced unexpected output:
 >     [java]
 > **********************************************************************
 >     [java] *** line 2 of expected output missing:
 >     [java] - socket.error
 >     [java]
 > **********************************************************************

This was very simple; the expected output file was expecting the 
test_socket module to print "socket.error", which it doesn't: I spent a 
lot of time to make sure that it doesn't!

I've checked in the correction to the Lib/test/output/test_socket file; 
that socket error will no longer appear.

Regards,

Alan.